Touch and Movement

1
Touch and Movement
2
Outline

• Review
• Touch
• Movement
• Fitts Law
• Steering Law
• Sloppy Selection Path

3
Touch/haptics

• Three sensors in skin
• Thermoreceptors
• Respond to changes in termperature
• Nociceptors
• Respond to intense heat, pressure or pain
• Mechanoreceptors
• Respond to pressure
• Mainly focus on mechanoreceptors, for obvious
  reasons
• The interface that injures is probably bad
• Two types of mechanoreceptors
• Rapidly adapting respond quickly as skin dented,
  but then stop
• Slowly adapting respond to continuous pressure
• E.g. force feedback joystick project, etc.
• Some areas are more sensitive
• Two point test

4
Touch/haptics

• Second aspect kinesthesis
• Awareness of the position of body and limbs.
• Ideas for use of this have been presented
• E.g., Jeff Raskin proposed using kinesthesis for
  mode selection holding down a button -gt
  selection mode with pen interfaces
• Three types
• Rapidly adapting when a limb is moved quickly
• Slowly adapting respond to movement and static
  position
• Positional receptors let you know where your
  body is

5
Tangible interfaces

• Under research topics
• User interfaces focused around touch and the
  movement and placement of physical objects
• Compelling
• Toy
• Called TUIs.
• Scott Klemmer working on toolkit

6
Movement

• Why is movement important in an interface?
• Answers?

7
Movement

• Why is movement important in an interface?
• Answers?
• Fitts Law
• Steering Law

8
Fitts Law

• Important to understand, as the formula provided
  a basis for trajectory movement studies
• Developed in the 50s by experimental
  psychologists
• Information theory to understand human
  perceptual, cognitive, and motor processes
• relationship found that models speed/accuracy
  tradeoffs in aimed movements

9
T a b log2 ( A / W c)

• T time needed to point to a target
• W width of target
• A distance to target
• a b empirically determined constants. c
  0, 0.5, or 1
• (different values used for c by different
  experimenters to achieve better fits for data,
  avoid negative values, etc.)

10
Key Components

• Difficulty of the motor task called the index of
  difficulty(ID). A larger ID more difficult
  task.
• Index of performance, to measure input device
  efficiency (IP).
• ID b log2 ( A / W c)
• IP ID / T or 1 / b

11
Physical interpretation of Fitts Law

• Larger objects at closer range acquired faster
  than smaller ones that are farther away
• Showed what we know intuitively the faster we
  move, the less precise our movements are

12
Advantages of Fitts Law

• Allowed experimenters to translate performance
  scores in different tasks (pointing/tapping) into
  a performance index
• Index of performance is independent of specific
  task parameters, so we can compare results across
  studies which use different experimental details

13
Disadvantages of Fitts Law

• describes only one type of movement
  (pointing/selection)
• modern computer input devices do more that point
  to targets. Based on trajectory input
• Need to be able to compare performance in such
  tasks as
• Pursuit tracking, free-hand inking, tracing, and
  constrained motion

14
Fitts Law Example
From Landays HCI slides Im still not sold on
Pie menus
15
Steering Law
16
Studies by Jonny Accot and Shumin Zhai

• 2 papers behavior of trajectory based tasks.
• 1997 development of formula and experiments to
  compare performance in such tasks with a single
  device (more mathematical.)
• 1999 application of the formula to several tests
  to compare different input devices

17
Goals

• Model motion mathematically
• Extension of fitts law
• Choose paradigm which focuses on steering between
  boundaries

18
Basis for Study of Trajectory Based Tasks

• Early handwriting studies show that time needed
  for writing a character constant, regardless of
  size of character
• but larger characters (larger amplitude) less
  precise
• Also showed that faster movements shown to be
  less precise
• So time to produce trajectories sets the relative
  speed-accuracy ratio
• New experiment devised to establish quantitative
  relationships between completion time and
  movement constraints in trajectory based tasks.

19
Experiments

• 1st experiment GOAL PASSING
• Subjects passed a stylus from one end to other as
  fast as possible. Several trials with different
  amplitudes and widths.
• Results show that goal passing follows the same
  law as Fitts tapping task. Graph shows linear
  relationship between ID and completion time.

20
Second Experiment Increasing Constraints

• Next add more constraints
• more goals placed on trajectory
• what happens when number of goals approaches
  infinity
• when there are no goals between end points
  equation is
• ID1 log2 (A/W 1)
• with N goals
• IDN log2 (A / NW 1)

21

• When N-gt infinity, the task approaches tunnel
  traveling
• index of difficulty can be determined by finding
  limit of index of difficulty recursion (IDN)
• Expand Taylor Series of
• log2 (x 1)
• To find
• ID8 lim IDN (A / W ln 2)
• N-gt 8
• So difficulty to achieve this task not related to
  log (A/W) but just (A/W), giving equation
• MT a b (A / W)

22
Third Experiment Narrowing Tunnel

• wanted to see if it could also apply when linear
  trajectory not in a constant path, something like
  this
• broke up the task into elemental steering tasks,
  and found sum of these
• integrating these paths shows that the index of
  difficulty is
• IDNT A / (W2 W1) ln (W2 / W1)
• Results index of difficulty forms a linear
  relationship. High error rate due to the high
  constraint on the right side of the tunnel (18)

23
Establishing a Global Law

• Applied narrowing tunnel concept of integrating
  the inverse of the path width along the
  trajectory to more complex paths to produce a
  generic formula
• Verified by plugging in known formulas (e.g. 2nd
  experiment formula where constraints were
  increased to tend towards infinity). Resulting
  formula is the same
• Tc a b (1/W )?c ds a b (A/W)

24
Fourth Experiment Spiral Tunnel

• Spiral tunnel used to test a complex path
• Subjects had to draw a line from center to end
• Parameters varied
• w width of spiral
• n number of turns taken
• experiments were run with 4 different values for
  each parameter

25
Results of Spiral Test

• the formula to predict ID also applies to this
  more complex case
• Proves that they derived a global law to predict
  the total time to perform a steering task
• Local law, to model instantaneous speed
• v(s) W(s)
• ?
• v(s) velocity of the limb, at the curvilinear
  point
• W(s) width of path at the same point
• ? empirically determined constant

26
Design Implications

• hierarchical menu item selection involves 2 or
  more linear path steering tasks. This can be
  modeled as follows
• Tn a b (nh/w) a b (w/h)
• 2a b ( n/x x) with x w/h
• h height of sub menu
• n submenu level
• So T is minimal when x vn or w vn h
• the number of menu items there are, the greater
  the quotient w/h is
• Can be used to compare designs, i.e. linear
  hierarchical menus and hierarchic pie menus

27
Justifification IBMs Trackpoint

• 5 input devices tested, in 2 steering tasks
  (linear and circular steering tasks)
• Linear representing the task of navigating in
  hierarchical menus
• Circular examines the ability to move along
  curved trajectories

28
Formulas used to find difficulty of steering
through a tunnel

• For linear tasks (straight tunnels)
• T a b (A / W)
• For circle tunnel steering tasks
• T a b (2 ? R / W)

29
Devices Used

• Logitech Mouseman
• Wacom ArtZ II 6 x 8 tablet and stylus
• Cirque Glidepoint touchpad 2

30

• Marcus trackball
• IBM trackpoint 3

31
Experiment Setup

• All software parameters left at default level
  rather than optimizing them (OK, since this is
  what the average user would do anyways just
  plug it in and use it)
• variables besides task type and device, movement
  amplitude and path width

32
Results

• All devices fit the steering model
• Device significantly affected the steering time
• Mouse and tablet outperformed all others
  significantly
• Circular task more difficult than linear one,
  even when the lengths and widths were the same in
  both cases
• Some devices better than others in one task and
  worse in the other

33
(No Transcript)
34
Effect of Variables

• In some cases, changed the relative performance
  of devices.
• Amplitude Larger amplitude Larger Steering
  time.
• The touchpad and trackball faster than trackpoint
  with small amplitudes, but slower with larger
  ones. (Need repeated strokes for large
  movements.)
• Width Larger width smaller steering time.
• trackball performed similarly to touchpad in
  wider tunnels, but was faster in narrow tunnel
  steering.

35
Rankings of devices
IP 1/b used to compute index of performance,
that indicates the steering time increase as a
function of task difficulty

• Circular steering
• Mouse (5.5)
• Tablet (5.4)
• Trackpoint (3.7)
• Trackball (3.0)
• Touchpad (2.5)

• Linear steering
• Tablet (14.4)
• Mouse (14.3)
• Trackpoint (8.7)
• Touchpad (6.7)
• Trackball (5.3)

36
Conclusions

• Performance can generally be grouped in this
  order
• Tablet and mouse
• Trackpoint
• Touchpad and trackball
• Points worth noting
• Mouse and tablet are mostly similar, but steering
  time of tablet shorter in circular or narrow
  tasks as there is higher dexterity associated
  with the tablet
• Relative performance of the trackball and
  touchpad switched order between linear and
  circular tasks
• Trackpoint more advantageous than the trackball
  or touchpad for longer tunnels since dont have
  to release and re-engage it

37

• Importantly, the steering law proved to hold for
  all the devices tested as none deviated more then
  0.02 from the law
• Most devices tested were not developed with the
  steering law paradigm as their guide
• Can be used to refine already existing ones, and
  to help develop new ones

38
The Sloppy Selection Path

• Sometimes users are vague in the expression of
  their intention
• For example
• I want to select this line of text now.
• Q Can we measure how accurately a user is
  expressing their intention?
• My research

39
Problem formulation
Given a Path
Can we infer a tolerance
40
Problem formulation
In a particular region of interest, whats the
probability a point outside the selection could
be inside it (or vice versa)?
Inside selection
41
HCI work Steering Law
Fitts Law generalized to constrained paths Data
tablet/mouse Geared toward pop-out menu navigation
42
Neuroscience Trajectory analysis
2/3 Power Law Minimum Jerk Law
A(t) K C(t)2/3
Tangential Speed
V(t) K R(t)1/3
Time
43
Results to date
44
Results to date
45
Current status

• Patent filed and pending
• Project is on-going, with a lot to do
• Need to do a lot of user trials to drive the
  development of a more accurate model
• Need to integrate it with grouping and some
  notion of strength of groups
• Need some intelligent way to display alternatives
• Need some interface work to study interaction
  with different options and selection among options