Eyedraw: Enabling Children With Severe Motor Impairments To Draw

    Drawing With Children. Eyedraw is a software program that, when run on a computer with an eye tracking device, enables children with severe motor disabilities to draw pictures by just ix.cs.uoregon.edu.

  • Id: # 046ddb4
  • File Type: Pdf
  • Author: ix.cs.uoregon.edu
  • File Size 3.9 MB
  • Read by User: 52 Times
  • Published: Monday, January 5, 2015
  • index: Drawing With Children


  • Read Online
EyeDraw: Enabling Children with Severe
Motor Impairments to Draw with Their Eyes
Anthony J. Hornof Anna Cavender
Computer and Information Science Computer Science and Engineering
University of Oregon University of Washington
Eugene, OR 97403 USA Seattle, WA 98195 USA
[email protected] [email protected]
EyeDraw is a software program that, when run on a
computer with an eye tracking device, enables children with
severe motor disabilities to draw pictures by just moving
their eyes. This paper discusses the motivation for building
the software, how the program works, the iterative
development of two versions of the software, user testing of
the two versions by people with and without disabilities, and
modifications to the software based on user testing.
Feedback from both children and adults with disabilities, and
from their caregivers, was especially helpful in the design
process. The project identifies challenges that are unique to
controlling a computer with the eyes, and unique to writing
software for children with severe motor impairments.
Categories & Subject Descriptors: H.5.2 [Information
Interfaces and Presentation]: User Interfaces - input devices
and strategies, interaction styles.
General Terms: Design, Human Factors.
Keywords: Art, children, drawing, eye tracking, input
devices, interaction techniques, universal access.
New software is needed to enable people to control their
computers with eye movements. This need is especially
acute for people with severely impaired motor abilities, who
cannot move their limbs or speak, such as people with partial
paralysis resulting from Amyotrophic Lateral Sclerosis
(ALS, or “Lou Gehrig’s disease”), brain injury, or cerebral
palsy. These people are severely limited in their ability to
interact and communicate with the rest of the world. Despite
these severe disabilities, many of these users retain normal
control of their eyes, which opens the door to perhaps the
best and most noninvasive means for these people to interact
and communicate with the world—with eye movements.
Overall, few software applications have been specifically
designed to be controlled with eye movements. Exceptions
include software for typing with the eyes by moving the gaze
across a keyboard displayed on the computer screen [8].
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that
copies bear this notice and the full citation on the first page. To copy
otherwise, or republish, to post on servers or to redistribute to lists,
requires prior specific permission and/or a fee.
CHI 2005, April 2–7, 2005, Portland, Oregon, USA.
Copyright 2005 ACM 1-58113-998-5/05/0004…$5.00.
However, eye-controlled software is not available for the
vast majority of the activities that people without disabilities
accomplish on their computers or with pencil and paper.
There is a particular need for new eye tracking software
applications to be developed for children. Children have
special interaction and communication needs that, if not met,
will impede their social, emotional, educational, and creative
development, and further reduce the ability of children with
complex physical disabilities to function in society.
EyeDraw is a software program that, when run on a
computer with an eye tracking device, enables children and
young adults with severe motor impairments to draw with
their eyes. EyeDraw is developed iteratively based on user
feedback. Previous work by the authors [4] shows how
analysis of human perceptual-motor control contributed to
the initial design of EyeDraw, and how children without
disabilities could use the tool. This paper shows EyeDraw
progressing from Version 1 to Version 2 based on user
studies, and presents the results of user observation studies
conducted with children and adults with disabilities who
successfully used EyeDraw to draw pictures with their eyes.
For over twenty years, researchers have been building
systems that use the eyes as a direct input to the computer [1,
5, 11]. There has been more recent interest in finding ways
to use eye position in some secondary, useful manner, such
as monitoring a user’s attention to find opportune times for
interruptions [10], or to jump the mouse cursor to the gaze
region when making manual mouse movements [12].
Overall, success has been limited in part because eye
tracking is technically challenging and labor-intensive, and
because eye movement data are noisy and difficult to
interpret [6]. However, there have also been some success
stories. Furthermore, improvements in the accuracy and
ease-of-use of eye trackers make it increasingly feasible to
build software applications tailored for eye control, such as
for drawing with the eyes.
It has been observed throughout the world that children
naturally progress through a series of qualitative stages when
learning to draw with paper and pencil: random scribble,
controlled scribble, basic forms, early pictorial, and later
pictorial [7]. Children follow the same stages of
development when learning to draw on computers [2].
Previous research suggests that important developmental
processes might be achieved through drawing with the eyes,
and provides a taxonomy and framework for analyzing the
drawings made using EyeDraw to determine if EyeDraw
supports the natural progression of learning to draw.
Previous approaches for drawing with the eyes use free-eye
drawing. In free-eye drawing, pixels on the screen are
colored-in wherever the eye tracker records the gaze on the
computer screen. Figure 1 shows free-eye drawing from
Tchalenko [9] and from EaglePaint [3]. Both systems have
produced drawings that would be categorized in the scribble
stages of drawing, but not in the basic forms or pictorial
stages [7]. Children have not used the systems to draw
recognizable objects and scenes such as people and houses.
Figure 1. On the left, three attempts to free-eye
draw the name “John” from Tchalenko [9]. On
the right, free-eye drawing in EaglePaint [3].
The difficulties in free-eye drawing can be explained in part
based on the characteristics of human visual perception and
oculomotor (eye movement) processing. First, free-eye
drawing jams together two task activities that are usually
independent when drawing a picture: eye movements to view
the drawing, and manual (hand) movements to draw lines.
Second, people do not have the same control over their eyes
as over their hands and other limbs. People can move their
eyes in short, quick bursts, but not slow adjusting
movements. EyeDraw accommodates these constraints.
Eye Tracking Terminology
To understand how EyeDraw works, it is useful to have a
basic understanding of how the eyes work, and how eye
trackers work. The gaze is the vector that goes from the eye
to the gazepoint, which is the point in a scene where a person
is looking. The gaze moves around a scene with a series of
quick jumps called saccades, each of which lasts roughly 30
ms. Between saccades, the gazepoint stays at the same
location (with a slight tremor) for a fixation that lasts
roughly 100 to 400 ms. A dwell is a long fixation. The
reason that the eyes move, in short, is so that people can put
items of interest into the high resolution vision which is at
the center of their gaze.
An eye tracker generally reports the gazepoint on the
computer screen 30 to 1000 times per second. EyeDraw uses
the LC Technologies Eyegaze eye tracker, which reports the
gazepoint 60 times per second, or once every 16.7 ms. The
system uses the pupil-center corneal-reflection technique.
EyeDraw averages the location of every six consecutive
gazepoints reported by the eye tracker and displays them on
the screen as the eye cursor. The eye cursor is a colored
square (seven pixels wide) that dances around the screen
wherever the user puts their eyes, with a small roughly 133
ms delay.
Alternating Between Looking and Drawing
EyeDraw enables the user, while keeping his or her gaze on
the picture, to shift between using their eyes to (a) just look
at the drawing and (b) add to the drawing. This smooth
subtask-switching is one of several differences between
EyeDraw and previous software for drawing with the eyes.
In both Tchalenko’s free-eye drawing system and in
EaglePaint, the ink effectively poured from the user’s gaze.
What resulted was a case of the “Midas touch” problem, in
which anything the user looked at became activated. The
user could not examine alternative spaces in which to draw
or pick up the pen to move to the next character without
putting down more ink all along the way. EyeDraw does not
have such a problem.
Figure 2 shows how a user controls the drawing process in
EyeDraw. The design departs from free-eye drawing by
providing control that is one level removed from the direct
coloring-in of pixels. Rather than drawing directly, the user
effectively manages a drawing process. It is somewhat
analogous to using a tool in drawing software for the general
public. The user defines the starting and ending point of a
line rather than drawing the line pixel by pixel. The
EyeDraw user is still, however, faced with the challenge of
using the visual modality to both determine where to place
the start and end of the line, and to place the points. This
problem, which is resolved by a tight control and feedback
loop around the eye cursor, is discussed next.
Figure 2. A storyboard showing how a user draws a
line in EyeDraw with eye movements and fixations.
State: Looking
User (U): The user is looking.
System (S): The system is
measuring dwell times.
Transition: A command is proposed.
U: The user holds gaze steady for 500 ms.
S: The system detects the dwell.
State: Drawing
U: The user noticed the cursor changed and
is deciding whether to commit the action.
S: The system is measuring the dwell time.
Transition Option 1: The action is completed.
U: The user holds the gaze steady for another 500 ms.
S: The system detects the dwell, commits the drawing
action, and emits an audible “click.”
Transition Option 2: The action is abandoned.
U: The user moves their eyes to a new location in < 500 ms.
S: The system resets for the next drawing command.
Figure 3. A state transition diagram of the two eye cursor states and the transitions between them when drawing.
Issuing Drawing Commands
This section describes how the user controls the state of the
eye cursor, and thus the drawing process. Figure 3 shows the
two states that the cursor moves through when the user
issues a drawing command. The first Looking state uses a
green cursor. As long as the user keeps moving their eyes
around, the cursor will stay green.
If the gaze dwells at a location for a minimum amount of
time, the program enters a Drawing state and the cursor
changes to red. The threshold is initially set to 500 ms, but is
adjustable to accommodate different levels of ability. To
stop the command from being issued, the user moves his or
her eyes from the current location within 500 ms. This
returns the user to the green Looking state without issuing a
command. To issue the command, the user continues
dwelling for another 500 ms, at which point EyeDraw
executes the drawing command. Auditory feedback also
confirms the drawing command was executed. The program
then automatically returns to the Looking state.
This transition between the looking and drawing states can
be applied to a wide variety of drawing tools, including a
line, square, and circle. The same basic control technique
can be used to position and “stamp” clip-art onto a drawing.
We have thus far developed two versions of EyeDraw.
Version 1 includes a minimum set of eye-control features:
tools for drawing lines and circles; an “undo” button; a grid
of dots to help the user dwell at a chosen location; and a
facility to save and retrieve drawings. To assist the
developers, the program also records all eye movements so
that they can be replayed later in the lab.
Version 1 was evaluated with two user observation
studies—a local study in which the software was evaluated
by children and adults without disabilities, and a remote
study in which the software was evaluated by adults with
severe motor disabilities.
Evaluation by Children Without Disabilities
Though the software is ultimately designed and intended for
children with disabilities, user observation studies using
children without disabilities are useful because this enables
us to evaluate the software with many users in a highly
controlled environment, and to have more extensive
discussions with the children about their experiences with
the software. Testing the software with children and adults
with disabilities who use an eye tracker to communicate is
critical, but also very difficult in part because these users are
widely distributed across the globe.
EyeDraw Version 1 was initially evaluated by children and
adults without disabilities. The primary question in our user
observation study was whether people could use EyeDraw to
draw recognizable pictures. Secondary questions included
(a) which parts of the drawing program were easier or harder
to use, (b) what were the preferred settings for issuing
drawing commands (for example, if 500 ms is a good dwell
threshold), and (c) what were the participants’ subjective
impressions of using the software.
Ten participants without disabilities were recruited. Four
were female and six were male. Half were children (under
eighteen years of age), with ages of 7, 10, 13, 14, and 16.
The other half were 21 to 36 years of age, with an average
age of 26.
Each session lasted a little under an hour. After preliminary
paperwork and a brief questionnaire, the eye tracker was
calibrated to the participant. Each participant was briefly
introduced to the basic functionality of the software, and
asked to make some drawings.
The preferred dwell time was consistently found to be 500
ms (we tried 250, 500, 750, and 1000 ms). The transition
from the green to red cursor to indicate the transition from
Looking to Drawing can optionally be set to include an
additional intermediary yellow cursor state. Users, however,
preferred the simpler two-state control.
Participants were asked to rate the ease of use and ease of
learning of the program on a scale from 1 to 4, with 1 as
“very easy” and 4 as “very hard.” Participants generally
found the program easy to learn (mean=1.6) and easy to use
(1.7). The easiest tasks were clicking on the buttons with the
eyes (1.2) and saving the drawings (1.3). The hardest tasks
were controlling the eye cursor (2.2) and controlling the
drawing (2.3). Participants found the grid of dots useful.
Seven out of the eight participants were able to draw a
picture that was judged by the authors to be of a clearly
recognizable scene. The youngest participant (seven years
old) was the only one who did not draw a full picture, though
he did gain control over the drawing process and was able to
follow an experimenter’s suggestions to draw lines from one
region of the screen to another.
Figure 4 shows two of the drawings made by children using
the EyeDraw system. The drawings can be identified as a
girl, and a house in the sun.
Figure 4. Drawings made by children without
disabilities using EyeDraw Version 1.
Local users identified a number of areas for improvement in
the software, but few major problems. For the most part,
local users helped us to verify the usability, adjust some
settings, and identify what features to add next.
Though most participants volunteered that they found the
activity to be fun, it was not particularly easy. For those
participants who created a second drawing, they liked it
better than the first. It appears as if drawing with the eyes
requires a great deal of focused attention, but that it gets
easier with practice.
Evaluation by the Target Audience
Given that the goal of EyeDraw is to provide children with
creative and developmental experiences that are otherwise
unavailable to them because of a severe motor impairment,
the ultimate test of EyeDraw is w hether children with severe
motor impairments can use the software to draw with their
eyes. Nonetheless, the software was evaluated by both
children and adults with disabilities in part to increase the
number of possible users, but also because the adults tend to
be more communicative. Even though they are nonverbal in
a conventional sense, many are “digitally verbal” with their
eye-controlled communication system. These users have
learned to advocate for themselves and thus by association
for younger users in a voice that the younger users have not
yet acquired. Nancy Cleveland, a registered nurse working
for LC Technologies, assisted in the recruitment of our users
with disabilities. She specifically put us in touch with one
young adult because she anticipated that this user could tell
us whether the software would have worked for her when
she was a child. This feedback turned out to be readily
Four “remote” users tested EyeDraw. We recruited
participants from a population of users of the LC
Technologies Eyegaze Communication System (the
“Eyegaze system”). Testing was conducted remotely with
the assistance of caregivers. All of the users have severe
cerebral palsy, and no functional use of their arms and legs.
All but one of the participants (User #2) are nonverbal. All
have little or no purposeful movement in their arms and legs,
but normal control of their eyes. Each remote test site also
had one primary caregiver who installed the software,
administered the user study, and reported the results. The
caregiver was either the user’s mother or an assistive
technology specialist at a care facility.
We refer to the users as User #1 through User #4. All four
are introduced here even though only the first two evaluated
Version 1. (All four tested Version 2, discussed later.)
User #1 is an 18-year-old woman who has used the Eyegaze
system at home for ten years. She is a “power user” of the
system, using her Eyegaze system up to 8 hours a day. She
typically works with two computers simultaneously, using
the Eyegaze system to move the mouse cursor and eye-type
on a second Windows computer, which runs regular desktop
software and is connected to the internet. The participant
uses the system to chat with friends, write poetry, surf the
internet, and do homework. The user shared numerous
emails with the authors during the course of the study. She
has tried in the past to paint with a head pointer, but found it
difficult because she had to move her head to one position to
paint, and to another position to see what she was painting;
the device also caused headaches.
User #2 is a 61-year-old man with a Master’s in
Rehabilitation Counseling. He is able to verbally express his
needs but his speech is slower than normal. He has been
using the Eyegaze system for a year and a half, for an
average of about an hour per day. He uses it to write letters
to family and friends, and articles for a newsletter associated
with the long term care facility where he resides.
User #3 is a 12-year-old boy who has used his Eyegaze
system at home and at school for about three years. He uses
it for typing, schoolwork, communication, playing computer
games, and getting on the internet, especially to visit sports
websites. The user has had little opportunity for visual
creative expression. His mother explains that he enjoys
“hand over hand” arts and crafts, in which a caregiver
completes the activity and puts the child’s hand through the
motions, “but they can be frustrating for him,” presumably
because he has no direct control over the activity.
User #4 is a 9-year-old boy who has used an eye tracker
along with Speaking Dynamically Pro since the age of 3.
His mother wishes that the eye tracker could also be used to
interact with storybooks, and with reading and math
Each caregiver was sent a packet that included the EyeDraw
software, instructions for installation and testing, consent
forms, and a set of questions. One set of questions asked
about the participant, such as how long the participant has
been using the Eyegaze system, the nature of his or her
motor impairments, and in what sorts of creative activities he
or she engages. Another set of questions, to be answered
after trying EyeDraw, asked about the usability and
learnability of the software. Open-ended questions included
the following: What were your overall impressions? How
was the drawing control? How was the overall control of the
program? How can we improve the overall functionality of
the program? Specific questions included but were not
limited to the following: Could you start and end the lines
where you wanted? Did the buttons sometimes get clicked
by accident? What new features would you like to see?
Almost every test site experienced some sort of initial
technical challenge such as requiring updated system files, or
hurdles in installing the EyeDraw software.
User #1 used EyeDraw Version 1 for seven sessions, for
about an hour per session, and saved 28 drawings. The
clearest feedback provided by this user was that she did not
want to use EyeDraw if it were not accessible through the
Eyegaze communication system’s main menu. For this user
to test EyeDraw, her caregiver had to quit out of the
communication system and manually start EyeDraw. The
user could later quit out of EyeDraw, but this did not
automatically return to the communication system. It instead
left the system inaccessible to the user. For the period of
time during which the user tested EyeDraw, she was isolated
from the control and independence afforded by the Eyegaze
system. This user found this to be entirely unacceptable.
Though she agreed to try the software, she wrote: “I want to
get to stuff on my own.”
Figure 5 shows two of the drawings created by User #1. She
described the drawings as specific scenes: “someone yelling”
and “someone trying to do the jumping jacks.” She reported
that she liked the drawings she created with EyeDraw, felt
that she was in control and could draw what she wanted, and
found nothing difficult or frustrating other than EyeDraw not
being in the main Eyegaze menu. She did report, though,
that sometimes the eye cursor was jerky and unstable, thus
making it hard to draw. She reported that EyeDraw was
easier to use than painting with a head pointer because she
could look at what she was drawing. She also reported that
the program felt “too slow because I couldn't start drawing
right away, I had to wait for the dot to change colors.”
Figure 5. User #1 drew these two pictures using
EyeDraw and described them as “someone yelling”
and “someone trying to do the jumping jacks.”
User #1 answered some of the specific questions about the
interface. She reported that the clicking sounds that
accompanied the transition from green to yellow to red were
helpful, that she did not use the grid very often, that the
buttons were easy to click on, and that “it’s easy to save” the
New features requested by the user included: putting
EyeDraw in the Eyegaze menu; a tool for drawing squares;
and adding color. She also requested “a text button where I
can type on my drawings. It also needs a spray-brush.”
User #2 used EyeDraw Version 1 for seven sessions, for an
average of 36 minutes per session. He typically produced no
drawings per session. User #2 had great difficulty using the
software. His four hours of working with the program
produced only a few drawings that had more than one line or
shape. Despite the difficulties, the caregiver reported that
the user was “really excited about the program.”
To try to understand the user’s difficulty, we replayed and
watched all of the user’s eye movement data at the lab. The
source of the difficulty appeared to be that the eye tracker
was not tracking his gaze smoothly. The eye cursor was
very jittery and erratic, which would make it very difficult to
issue the gaze-based drawing commands to start and end
lines and circles. The caregiver decreased the fixation-
detection time from the default 500 ms to the fastest possible
250 ms in the first session, and kept it at that setting for
subsequent sessions. Even at this setting, the user had a
difficult time issuing line-drawing commands. On the
occasion that he did draw the lines and circles, they appeared
so quickly that he did not appear to be in control. Even with
the grid turned on, and the user clearly trying to fixate the
dots, the cursor was generally too jittery to issue a command.
The caregiver pointed out to us that an eye image appeared
on the screen in all other Eyegaze software, but not ours.
This is an image of the eye as seen by the eye tracking
camera, along with a color-coding that indicates if the eye is
currently too close (red) or too far (green). She suggested
that perhaps the difficulty in tracking resulted in part because
the user could not see when he was in and out of optimal
range of the camera. If he could see the image, he could
make small head and neck adjustments to get the eyes back
in the optimal range. She suggested that we add the eye
image to the EyeDraw screen.
Other suggestions from the user and caregiver included
adding: colors, designs, patterns, and textures; different
sounds for the changing of the eye cursor from green to
yellow to red; enhanced eye-controlled tool bars and menus;
and coloring book exercises. They also suggested that the
eye-controlled buttons “speak” their function when the eyes
look at them.
Remote users, unlike the local users, identified two major
problems with EyeDraw that would hinder EyeDraw’s
usefulness to our target population. The problems included
the lack of an eye image on the screen, and EyeDraw not
being in the Eyegaze system’s main eye-controlled menu.
These are important fundamental problems that were readily
identified by adult users with disabilities, and for which it
was not necessary to work with children with disabilities.
To address User #2’s difficulties with the jittery cursor, we
added a feature so that the end-user could adjust the spatial
distribution of the fixation-detection algorithm. The feature
is the complement to how the user can already adjust the
dwell time of the algorithm. We replayed User #2’s jittery
eye movement data after increasing the spatial distribution
from the standard 0.25 inches to 0.75 inches, and this
resulted in many more fixation commands being recognized.
Requiring caregiver intervention to start EyeDraw was
frustrating for User #1. However, it is difficult to add
EyeDraw to the main system menu in part because this
requires us to modify the software that these participants use
to communicate with the world, and we are reluctant to risk
introducing bugs. Nonetheless, it is clearly important to
make EyeDraw and other software for this population
accessible within their current eye-controlled environment.
User #1 clearly used the program to draw. She was
reasonably happy with her drawings. Taking a narrow view
of what drawings should look like, the drawings are perhaps
not “perfect,” but then again neither are the drawings made
by anyone learning to draw for the first time. User #1’s
drawings are not immediately recognizable as specific things
or scenes, but they do appear to be deliberately organized,
composed, and drawn. The drawings look like the emergent
diagram shapes seen in the basic forms stage of learning to
draw [7]. It appears as if the user was trying to draw faces,
which typically appear in the early pictorial stage of
drawing. When the user explained the content of the
drawings, the intentionality was readily apparent. It appears
as if the user was able to enter one of the intermediary stages
in the natural progression of learning to draw. The software
clearly supports more than just the early scribbling stages of
drawing, which may be the extent to which free-eye drawing
can be used.
EyeDraw Version 1 was extended to Version 2 based on user
observations, user feedback, watching children work on art
projects, and the literature on children’s drawing. Existing
features were refined to improve usability of the basic eye-
drawing functionality. New features were also added.
Figure 6 shows a screenshot of Version 2. Version 2 has the
same features as Version 1 plus: (a) a display of the camera
image in the top right of the screen, to help the user stay in
range, (b) user-defined settings that provide the user with
various controls (shown in the middle of Figure 6) such as
the dwell detection thresholds, and (c) audio feedback that
reports the current state of the eye cursor when drawing—a
different note is played for each transition.
Version 2 also adds rectangle- and polygon-drawing tools;
colors; stamps that can be placed with a dwell; and a “Dot
On/Dot Off” button that parks the eye cursor and lets the
user look around without accidentally issuing commands.
Figure 6. A screenshot of EyeDraw Version 2 with
the eye-controlled Settings dialog box opened.
Version 2 was evaluated with two user observation
studies—a local study in which the software was evaluated
by children without disabilities, and a remote study in which
the software was evaluated by children and adults with
severe motor disabilities.
Evaluation by Children Without Disabilities
Children without disabilities evaluated EyeDraw Version 2.
The two driving research questions in the evaluation were
(a) whether the users would take advantage of the new
features to draw more “interesting” pictures and (b) whether
the refinements to the basic control system would make
EyeDraw easier to use.
All twelve participants were children without disabilities
between the ages of 6 and 14, with an average age of 10.
Seven were male and five were female. Two users had
evaluated Version 1 and are considered “longitudinal users.”
Ten were new users who had never before controlled a
computer with their eyes.
Each session lasted about one hour. Seven of the
participants also returned for a second hour-long session at a
later day, typically two weeks later. After preliminary paper
work and a brief questionnaire, the eye tracker was calibrated
to the participant. The participants were then given three to
five minutes of free time to explore the software. For the
new users, this time was spent learning how the program
works and how to draw with it. For the two longitudinal
users, this time was spent getting acquainted with the new
features and changes made to the program. For the next four
to six minutes the participants were asked to experiment with
different user-defined settings in order to find those that were
most preferable. Users were specifically asked if they found
the audio feedback to be useful. If the answer was no, they
were asked if the noise was really not useful or if the noise
was just annoying. If they confirmed that it was really not
useful, the sound was turned off. Otherwise, it was left on.
The remainder of the study consisted of two drawing
sessions, about fifteen minutes each, and a playback session,
usually about ten minutes. During the drawing sessions, the
user was told they could draw anything they desired.
Midway through each drawing session the sound was turned
back on (if currently off) and the user was asked to
reevaluate all settings including the sound.
For the playback session, the user’s drawing sessions were
replayed for the user at about four times the original speed.
This gave the users a chance to talk about their drawing
experience as it unfolded; it is generally difficult to talk and
draw with the eyes at the same time. Lastly, the participants
filled out a post-experiment questionnaire about the program.
As mentioned, most of the users returned for another one
hour drawing session. Upon returning, users were not given
time limits for their drawings, but were encouraged to draw
as much as they liked. Users were presented with the same
post-experiment questionnaire as in the first visit.
In our local usability study of Version 1, we found the
preferred dwell time to be consistently 500 ms. This time we
narrowed the choices (to 300, 400, 500, and 600 ms) and
found no clearly preferred dwell time. In our previous study
we found that most users preferred the two state control
(green->red) rather than with the intermediary yellow cursor
state (green->yellow->red). This time we found that only
50% of users preferred the two-state control. As one user
commented, “I liked the three noises better because it
sounded more balanced, but I chose the two noises because I
wanted to go faster.” In general, the children reported that
the system was easy to use and learn; that it was initially
difficult to control the drawing process; and that it was easier
to control the drawing process on the return visit.
Figure 7 shows two drawings from the user study. All
twelve of the participants were able to draw with their eyes.
The authors attempted to categorize each of the 53 drawings
made based on the stages of learning to draw and found the
following results: 5 were determined to be in controlled
scribble stage, 13 in the basics forms stage, 11 in early
pictorial, and 21 in later pictorial. One of the drawings was
considered a “non-drawing” as the user simply spent time
learning the tool without drawing anything more than a few
stamps placed on the screen in meaningless order. None of
the drawings fit into the random scribble stage as all users
exhibited at least some control of their markings and placed
objects in deliberate arrangements.
Figure 7. Drawings made by children without
disabilities using EyeDraw Version 2.
Overall, the children found the software easy to use and easy
to learn, though the drawing control was initially tough.
Initially, the children were somewhat captivated, or perhaps
distracted, as they explored the wide range of features. The
two longitudinal users seemed to be more captivated by the
use of color in the basic lines and shapes which they learned
how to use in Version 1. The first-time users were more
taken by the stamps, even to the point of using stamps of
eyes that they found after laboriously drawing eyes with the
line tools, as in the second drawing in Figure 7.
The infatuation with the many features and the attraction to
the stamps permitted less time for the creation of a drawing,
prompted us to invite all of the children back for a second
session to see if the novelty of the many features would wear
off and the children would spend more time drawing. They
did. In the second session, the children were much more
engaged with creating images. The drawings started to
develop in pictorial quality.
Evaluation by the Target Audience
As with EyeDraw Version 1, the ultimate test of Version 2 is
whether it can be used by children with severe motor
impairments. The software was evaluated by both children
and adults who communicate using an eye tracker.
Participants and Procedure
All four remote users that were introduced earlier (Users #1
through #4) evaluated EyeDraw Version 2. The procedure
was nearly identical to that for evaluating Version 1. One
additional step was that some of the drawings were sent back
to the users with a request for eye-typed comments, with
questions such as: Do you remember drawing this? Do you
like the drawing? Is it a picture of something special, or
were you just sort of practicing? If it is a picture of
something special, can you tell me what it is?
All four remote users were able to use EyeDraw Version 2,
but had a wide range of success in terms of its ease of use.
User #1 (18-year-old woman) tested Version 2, and saved
fifteen unique drawings in five sessions. F igure 8 shows her
using the system, and two of the drawings that she produced.
The user provided comments on the two drawings at the
bottom of Figure 8. About the left drawing, she eye-typed
“Here’s an ocean that I made up in my mind. I use the
stamps for the animals. The picture came out good.” About
the right drawing, she confirmed that it is a landscape and
wrote that she “did an awesome job”.
Both drawings clearly capture intentional, recognizable
scenes. The first also captures an inspired use of stamps to
create an abstract visual mass, a three-dimensional
representation of space that was specifically noted and
intended by the user. The drawings are somewhere between
the basic forms and early pictorial stages, which is further
along than her drawings from Version 1. It appea rs as if she
may be progressing through the stages of learning to draw
even in the course of these user observation studies.
Figure 8. User #1 using EyeDraw Version 2
(on the left of the two computer monitors),
and two drawings that she produced.
As with most users of Version 2, both local and remote, this
user was also immediately drawn to the new stamp feature.
She initially used them as word-like icons, to tell a little
story about how she loves her dog. She then used them more
as a drawing element, to create visual textures. During this
time, she did not use the line tools very much. But then she
slowly returned to the line tool, using it together with the
stamps. She did produce a couple of drawings with just lines
during this period, including one that she described as her
“bulldog/pug”. Her final drawing (Figure 8, bottom right)
integrates the shape and line tools with the stamps, for a
somewhat complex visual integration.
User #2 (61-year-old man) found EyeDraw Version 2 easier
to use than Version 1. He used EyeDraw Version 2 for just
three sessions. The first two sessions were about fifteen
minutes each, during which he was able to put down a few
more shapes and lines than in Version 1, but the drawings
were still very sparse, as with Version 1. In his third session
with EyeDraw 2, however, he experienced a breakthrough.
As reported by his caregiver, “He finally really got it.” This
single session lasted almost two hours, and produced sixteen
drawings, all of which were intermediary saves of one long
drawing. Figure 9 shows the resulting composite image,
which perhaps falls between the stages of controlled scribble
and basic forms. User #2 was now able to place lines,
shapes, and stamps. His early drawings were typically blank
or had a single line or shape. The new drawing demonstrates
greatly improved ability to control the drawing process.
Figure 9. A drawing by User #2.
User #2’s eye-control improved dramatically even though he
did not change the user-adjustable dispersion-threshold
parameter in the fixation-detection algorithm, a new user
control added to Version 2. The caregiver reported that the
user was finally able to use the program because (a) the eye
image on the screen made it much easier for him to stay in
range and be tracked accurately and (b) all of his practice
finally paid off, and he finally “got the concept” of how
EyeDraw works. As she reported, the final long session
“was unbelievable to me, and to him.”
User #3 (12-year-old boy) only tested Version 2, and saved
five drawings from three sessions. Figure 10 shows User #3
using EyeDraw, and two of the drawings that he produced.
The eye image, barely legible in the upper right of the
screen, appeared to be essential for this user to be able to use
the system. The user appeared to have control over the
software, though he seemed to have some trouble with the
various dialog boxes used to open and save drawings.
Watching a video taped by his mother, and also based on her
observations, the user seemed to get a little lost in some
dialog boxes, not knowing for sure what to do. It seemed as
if he accidentally deleted some drawings, clicking on “No
Save” when his mother was asking him to save his drawings.
Perhaps he really did not want to save it, but his actions did
not appear to be entirely deliberate.
Figure 10. User #3 using EyeDraw, and two
drawings that he produced. He described the
two drawings as (a) “bed” and (b) “home.”
Similarly, when ending a line, he sometimes went off the
drawing and onto the adjacent buttons, thus turning a square
into a circle, or clicking on “Save” and thus losing the shape
because he did not anchor the second corner. Local users
reported that the first of these two behaviors was particularly
annoying. The feedback was an example of how testing the
software with children without disabilities helped us to better
understand problems that a child with disabilities also had
but did not report.
Figure 10 shows two of User #3’s drawings, made during his
second and third sessions using EyeDraw. The drawings are
not immediately recognizable as scenes or objects, as were
the drawings made by children without disabilities, but they
demonstrate control of the tool and are meaningful to the
user. He described them as “bed” and “home.” The
drawings would probably be classified somewhere within the
controlled scribble and basic shapes stages. There appears to
be an effort to put marks in specific regions, and the user’s
comments also suggest some intentionality. Nonetheless,
perhaps some of the markings are like the scribbles children
make as they are learning how to hold a pencil and move
their arm. User #3 generally liked his drawings. He reported
that the program was hard to use but that it got easier with
User #4 (9-year-old boy) only tested Version 2, so he did not
benefit from earlier practice with Version 1. He only tried
Version 2 a few times, and for very brief sessions. He made
few drawings, and they tended to be very sparse.
The remote users had more success at using Version 2 than
Version 1. They were able to create lines and designs, and to
save their drawings. They appear to be successfully placing
visual objects at desired locations. Two users specifically
described the pictorial content of their drawings.
As we studied the drawings that were made by our users with
disabilities, we adjusted our criteria for the kinds of drawings
that would be needed to confirm that the software works for
this user group. We initially expected to see pictorial
drawings of the sort produced by our local users. We came
to realize that such drawings would probably take a long
time for our remote users to produce. The remote users had
not yet developed manual drawing skills as had our local
users. The remote users’ drawings are perhaps in some ways
akin to the drawings made by children as they figure out how
to control a crayon in their hand. However, our 18-year-old
remote user demonstrated some impressive pictorial skill,
and it appears as if her drawing ability might have developed
even within the course of the study.
Though User #3 seemed to have a very short attention span
with EyeDraw communication system, when visiting this
user, we noticed that he did have a lot of patience using the
Eyegaze system to control the mouse movement on a second
computer that was running the children’s software Backyard
Sports. The process was slow and tedious compared to a
typical pace, but nonetheless clearly gave the user much
satisfaction. Perhaps EyeDraw just needs to be more fun.
Perhaps it needs to also engage children without disabilities
in art activities, so that using EyeDraw becomes a peer-
encouraged social activity, just like playing Backyard Sports.
Though the eye tracker appeared to track User #3 accurately
and smoothly, he still had trouble drawing at first. Perhaps
our basic eye-command technique is not obvious at first.
Also, time on task is clearly critical. Overall, the user spent
relatively little time learning the system, at least compared to
our two older users. Clearly, EyeDraw needs to draw the
user in, perhaps using play and entertainment techniques
such as those used in children’s games.
Overall, remote testing of EyeDraw by children and adults
with severe motor impairments demonstrates that we have
successfully built and deployed a tool that can be used by
children and adults to draw with their eyes. Even though the
program is somewhat difficult to use at first, it gets much
easier with practice.
Across four remote users, we can identify two kinds of users
with two kinds of usage patterns. The two younger
participants use the eye tracker for relatively short periods of
time with somewhat constant caregiver attention. The two
older participants use it for longer sessions without a
caregiver, and evidently enjoy this independence. The
younger participants seemed to have less patience with
EyeDraw, whereas the two older participants seemed to have
more patience, in one case even trying it for several hours
before finally being able to draw.
We are particularly interested in addressing the needs of the
younger users, who seemed impatient and perhaps frustrated
when they were not able to draw, and who seemed to spend a
lot of time exploring features rather than drawing, at least
EyeDraw needs to help children to enjoy drawing as quickly
as possible. We noticed that the assistive technology
specialist and mother that we visited gave their children
continual constructive feedback and encouragement as the
child progressed through an activity. Though we disparaged
free-eye drawing earlier in our design process as a
problematic way to draw with the eyes, it may be appropriate
to limit the user to free-eye drawing when first using
EyeDraw. This would insure immediate positive feedback
even at the expense of control. New features, such as the
ability to change the color of the ink, or start and stop the
flow of ink, could be gradually introduced.
The data provide some insight for how and when EyeDraw
might gradually introduce new features to best support the
creative process. First, recall that the two less-patient
younger users tended to use the program for very short
periods of time, and with a caregiver continually present.
The caregivers had a very strong sense of what the children
would be able to do, and what would be too hard. For these
users, it would seem appropriate for the caregiver to control
the sequential introduction of new features. Second, recall
that our two older, more-patient users were willing to invest
a lot of time even though success was not immediate. We
suspect that they could handle all of the features up front and
still stay focused on learning to draw with their eyes.
We are currently developing Version 3, which will feature a
further-refined feature set as well as a progressive revealing
of increasingly advanced functionality. We are exploring
options for providing more immediate feedback and
encouragement to engage the children at every step. Some
of our target users had difficulty generating the first set of
dwells necessary to draw the first line and thus could not
experience and learn the basic control technique. Until using
EyeDraw, there has been little need for the users to make
successive dwells at the same location.
Having EyeDraw start the user with free-eye drawing might
give children a great deal of fun and satisfaction, and help
them to gain confidence that they can draw with their eyes.
Despite the difficulties associated with free-eye drawing, the
immediate feedback provided by the technique may help to
communicate to users the concept of using the eye position
to draw lines. Adding the ability to turn the ink off (with a
dwell) might also make free-eye drawing more tenable by
removing the “Midas touch” problem. The very eye-drawing
technique that was initially dismissed in our design process
might actually be the best introduction to the more enabling
but more complex drawing techniques built into EyeDraw.
EyeDraw is a software program that enables children and
adults with severe motor impairments to draw pictures by
moving their eyes. The software supports the range the
stages observed in the natural progression of learning to
draw. Children and adults, both with and without
disabilities, successfully used the software to produce
drawings that fell into the stages that may be appropriate
given each person’s previous experience with the activity. It
appears that EyeDraw may support the natural human
developmental pattern of learning to draw.
This research demonstrates how a detailed analysis and
understanding of fundamental human-perceptual constraints
and oculomotor control and feedback capabilities can be
applied to create human-computer interfaces that enable new
eye-control of software applications. These applications can
support open-ended creative processes such as that of visual
artistic composition to enable people with severe disabilities
who are currently locked out of fundamental human creative
and expressive opportunities to experience more of what life
has to offer.
Rob Hoselton contributed to the development and testing of
EyeDraw. Tim Halverson assisted with eye tracking and
design. Nancy Cleveland, Dixon Cleveland, and Peter
Norloff, of LC Technologies, provided technical and domain
expertise, and helped recruit participants.
The National Science Foundation supported this work
through a Research Experiences for Undergraduates
supplement to Grant IIS-0308244. The eye tracking lab used
for this project was supported by the Office of Naval
Research through Grant N00014-02-10440. Both of these
grants were awarded to the University of Oregon with
Anthony Hornof as the principal investigator.
1. Bolt, R. A. (1982). Eyes at the interface. Proceedings of
Human Factors in Computing Systems, New York:
ACM, 360-362.
2. Escobedo, T. H., & Bhargava, A. (1991). A study of
children's computer-generated graphics. Journal of
Computing in Childhood Education, 2(4), 3-25.
3. Gips, J., & Olivieri, P. (1996). EagleEyes: An Eye
Control System for Persons with Disabilities. The
Eleventh International Conference on Technology and
Persons with Disabilities.
4. Hornof, A. J., Cavender, A., & Hoselton, R. (2004).
EyeDraw: A system for drawing pictures with eye
movements. Proceedings of ASSETS 2004: The Sixth
International ACM SIGACCESS Conference on
Computers and Accessibility, New York: ACM, 86-93.
5. Jacob, R. J. K. (1990). What you look at is what you get:
Eye movement-based interaction techniques.
Proceedings of ACM CHI ‘90: Conference on Human
Factors in Computing Systems, ACM: New York, 11-18.
6. Jacob, R. J. K., & Karn, K. S. (2003). Eye tracking in
human-computer interaction and usability research:
Ready to deliver the promises (Section commentary). In
J. Hyona, R. Radach, & H. Deubel (Eds.), The Mind's
Eyes: Cognitive and Applied Aspects of Eye Movements.
Oxford: Elsevier Science.
7. Kellogg, R. (1970). Analyzing Children's Art. Palo Alto,
Calif., Mayfield Pub. Co.
8. Majaranta, P., & Räihä, K.-J. (2002). Twenty years of
eye typing: systems and design issue. Proceedings of the
Symposium on Eye Tracking Research and Applications:
ETRA 2002, New York: ACM Press, 15 - 22.
9. Tchalenko, J. (2001). Free-eye drawing. Point: Art and
Design Research Journal, 11, 36-41.
10. Vertegaal, R. (2003). “Attentive User Interfaces”
Editorial, Special Issue on Attentive User Interfaces.
Communications of the ACM, 46(3).
11. Ware, C., & Mikaelian, H. (1987). An evaluation of an
eye tracker as a device for computer input. Proceedings
of ACM CHI+GI, New York: ACM, 183-188.
12. Zhai, S., Morimoto, C., & Ihde, S. (1999). Manual and
gaze input cascaded (MAGIC) pointing. Proceedings of
ACM CHI ‘99: Conference on Human Factors in
Computing Systems, New York: ACM, 246-25.