The document describes a study that investigated using finger contact size as a way to simulate pressure levels for interacting with mobile devices. The study aimed to determine how many discrete pressure levels users could distinguish when touching a tablet screen with their index finger. It tested this with different trial types that either provided visual feedback of contact size or did not, and involved targets of varying difficulty. The results found that users were most accurate with trials that provided continuous feedback, and could distinguish up to 8 pressure levels. Trials without feedback performed best when targets were divided into 3 levels. The study provides insights into using contact size as a new dimension for touchscreen input.
Fat Finger - Use of contact Size as a form of simulated pressure.
1. Fat Finger
Using contact size for interacting with mobile
devices
Evangelos Tzemis
Supervisor: Sebastian Boring
Department of Computer Science
University of Copenhagen
Denmark
6th January 2015
1
6. Motivation
• How can we improve the way we interact with
mobile devices?
Vast evolution over the years.
7. 2D Interaction Approach
Despite this evolution, the main interaction principle
has, more or less, remained the same.
“Interaction is 2 Dimensional”.
Multitouch, combines multiple fingers
• Increments the expressiveness.
• Stacked on 2D principle.
8. Fat Finger Interaction Technique
Use finger’s contact size as a form of simulated
pressure.
Position(x, y) + Contact Size
3 Dimensional Interaction
Other approaches:
Styluses, pressure sensitive screens, etc.
9. Objective - Research Question
Understand the way Touch works
to apply
3D input on tablet-mobile devices
Research Question:
”To which extend are we able to distinguish the
different simulated pressure levels produced by our
fingers using a tablet device?”
11. Relevant Work
Herot and Weinzapfel [1978].“Touch and pressure sensing
open a rich channel for immediate and multi-dimensional
interaction”
Pressure widgets: 6 levels is the optimal division of the
pressure space.
Fat Thumb: adds dimension to input, seamless mode
switching.
”Contact size as an input parameter is closely related to
pressure (i.e., more pressure suggests a larger contact size
due to flattening of the finger” [Boring et al.]
12. Fat Finger Establishment
• Q1: How many discrete pressure levels are we able to
distinguish when using the index finger to interact with a tablet
device?
• Q2: In which region is our finger more capable to operate on?
• Q3: Which is the role and performance effect of visual
feedback?
• Q4: Does training in Fat Finger affect performance? Which is
the Learning Curve?
• Q5: In which detail are users able to develop haptic memory
on various pressure levels? Is it even possible?
14. Fundamental Principle
“Whenever a finger is touching the screen we should
be able to calculate the contact area between the
finger and the screen “
NSValue *val = [touch valueForKey:‘‘_pathMajorRadius’’];
float size = [val floatValue]; //size in pixels
Drawbacks:
• We can not obtain the exact contact size.
• Monitoring is not continuous but touch driven.
15. Basic Structure
• Target Selection Tasks
• Varying Type & Difficulty
Basic Interface:
• Targets placeholder
• Feedback placeholder
• Touch prompted region
Calibration:
• Measure minimum and
maximum desired contact size.
16. Feedback & Discrete Targeting - FD
Targets Region:
• Divided in N buckets.
• Target is Red.
Feedback Region:
• Visible & corresponds to the
contact area of our finger.
Mission:
Keep the edge of the blue region
inside the target for at least 1
second.
17. Feedback & Continuous Targeting - FC
Targets Region:
• Target is Red & placed inside
bucket.
• Yellow lines give offset of +-2%.
Feedback Region:
• Visible & corresponds to the
contact area of our finger.
Mission:
Keep the edge of the blue region
inside the yellow lines for at least
1 second.
18. No Feedback & Discrete Targeting - NFD
Targets Region:
• Divided in N buckets.
• Target is Red.
Feedback Region:
• Visible only after confirming
selection.
Mission:
Predict the appropriate contact
size. Lift your finger to confirm
selection.
19. No Feedback & Continuous Targeting -
NFC
Targets Region:
• Target is Red & placed
inside bucket.
Feedback Region:
• Visible only after confirming
selection.
Mission:
Predict the appropriate
contact size. Lift your finger to
confirm selection.
20. Repetition Structure
• N = [2, 3, 4, 6, 8,12,16]
• Each bucket will act as
a target.
• 4 different Categories.
• Randomise order.
• Repetition has 204
trials.
21. Final Work-Flow
• 3 repetitions of 204 unique trials each.
• Randomised order of trials in each repetition.
• 1st repetition acts as the Learning Phase.
• Between trials a Next Trial button.
• Confirmation sound when target has been hit
22. Data Manipulation - Monitoring
Variable Monitored in
User ID ALL
Trial ID
ALL
Type ID ALL
Repetition ID ALL
Min ALL
Max ALL
N ALL
Target ALL
RawInputValue ALL
Re Entries ALL
Re Touches ALL
Total Time ALL
Offset FC, NFD, NFC
Target Position FC, NFC
26. Participants
• 26 participants (+2 pilot users).
• Age [19 - 52]. Majority [21-30].
• 92% right-handed.
• 65% male.
• More experienced in touch-based devices than in
tablets.
• Received gift as compensation for their time and
effort.
27. Experiment Sequence
• Verbal instructions
• Demographic Information
• Calibration
• Experiment
• Assessment for each type of trial.
28. Hypotheses
• (H1). Feedback trials outperform No Feedback ones in offset.
• (H2). NFD outperforms NFC in terms of offset.
• (H3). CT gradually decreased over time.
• (H4). Error gradually decreased over time.
• (H5). CT when feedback is provided, is N dependent.
• (H6). Average contact areas will be subconsciously preferred.
• (H7). FD is the most preferred type of trial.
30. Task Completion Time
4 ∗ 7 (TypeID ∗ N) within subjects ANOVA
Feedback:
• CT dependent on the
target size.
• CT statistically
independent for N values
up to 8 buckets.
No Feedback:
• CT independent of N, also
in FC.
31. Offset
• TypeID, N, Combination do
not affect Offset.
• FC error rates always very
small. Sudden increase on
the last bucket.
• NFD and NFC error rates
follow a common pattern.
FC: mean=1,1%, std=0,1%
NFD & NFC identical.
Mean=16,1% & Std=1,1%.
3 ∗ 7 (TypeID ∗ N) within subjects ANOVA
32. Learning Curve - Total Time
Learning effect takes
place especially for
Repetition 1 & 2.
Only FC increases CT in
Repetition 3. Decreased
offsets.
Mental & Physical fatigue
might have influenced
performance.
33. Learning Curve - Offset
FC - NFD - NFC
Learning effect exists but
less obvious.
Combined with CT, we
indeed have a very
promising learning factor.
Certain error values are
affordable in NFD.
34. Re-Entries - Re-Touches
Re-Entries:
Feedback: Re-Entries N dependent. FC higher values
than FD.
No Feedback: 0.5 per trial (QuickRelease), N
independent.
Re-Touches:
FD: Only 0.6 per trial FC: 2 per trial.
Difficulties in targets placed close to the limits.
36. Discussion (1)
(H1). Feedback trials outperform No Feedback ones in offset.
Holds. In NF no indication on the feedback line, resulted in much higher
errors. Haptic memory developed was not that strong.
(H2). NFD outperforms NFC in terms of offset.
Holds. Similar error results. Certain error values are affordable in NFD,
and always higher than in NFC.
(H3). CT gradually decreased over time.
Holds. CT Learning curve depicts that. For FC, minor increase but does
not affect overall. Also caused on fatigue.
(H4). Error gradually decreased over time.
Holds in general. Especially in Repetition 2. Minor increments in Repetion
3 that also caused on fatigue.
37. Discussion (2)
(H5). CT when feedback is provided, is N dependent.
Does not hold. Not for FC, as remains stable. Alternatively, is
target-size dependent.
(H6). Average contact areas will be subconsciously
preferred.
Holds. Offset pattern in NF trials. Can not extract results
from feedback trials.
(H7). FD is the most preferred type of trial.
Holds. Rated by users from the assessment. Results also
depict that, even in Completion time.
38. Final Observations
Feedback Supplied:
Eight (8) levels is the optimal division of the contact-
size space.
No Feedback Supplied:
Three (3) levels is the optimal division. Also H6
influences precision. Also the restricted ability to
develop advanced haptic memory.
39. Next Steps…
Integrate into on market products.
Design of appropriate widgets.
Multi finger & Gesture integration.
42. Further Questions
• How many contact-size levels are we able to determine?
• How can we test that the N levels we managed to
determine are useful and distinguishable by the user?
43. Pressure on Mobile Devices (2)
Hardware augmented:
GraspZoom: pressure to the back of the device. Switch zoom to pan.
PressureText: use pressure-sensitive keypad.
Clarkson et al.: addition of simple pressure sensors under the keypad.
Software augmented:
ForceTap: infer strong vs gentle taps using accelerator
PseudoButton: use the built-in mic to emulate pressure. 5 pressure
levels. 94% accuracy. Similar to MicPen.
VibPress: detect pressure by measuring the level of vibration
absorption (accelerometer).
44. Contact shapes & Simulated pressure
Lee et al. [1985] touch-sensitive tablet prototype. Detect the on-
screen position & contact size of multiple fingers.
ShapeTouch: Utilises contact shape. Infers virtual contact forces
from contact regions. Enables interaction with virtual objects.
Fat Thumb: adds dimension to input, seamless mode switching.
”Contact size as an input parameter is closely related to
pressure (i.e., more pressure suggests a larger contact size due
to flattening of the finger” [Boring et al.]
Thumb Rock: In-drag gesture. Rolling the thumb back and forth
on a touch interface.
Buxton et al.
touch-sensitive tablet input
Pressure Widgets:
-Explores pressure-sensing capabilities of styluses.
-Number of distinguishable pressure levels
-Mechanisms to confirm target selection
-Impact of training
Utilise the contact area of our finger.
it gives the major radius of the touch in millimeters.
In the GSEvent, which is a lower-level representation of UIEvent, there is a structure known as GSPathInfo with members:
ios8: https://developer.apple.com/library/prerelease/ios/documentation/UIKit/Reference/UITouch_Class/index.html
Divided in N invisible buckets.
Size independent of N
Divided in N invisible buckets.
Size independent of N
- Observe and study the learning curve.
H6. this is the natural position of the finger.
H7. it best combines speed and accuracy.
No Removal of outliers.
We use ANOVA to determine if the mean values for offset are statistically different.
Mean FD = 2.387s
Mean FC = 5.485s
Mean NFC = 1.237s
Mean NFD = 1.365s
Of course, we can export basic information about the means by just comparing them, but we want to know what do these directional differences in the means infer about our results.
Analysis of variance (ANOVA) is a collection of statistical models used in order to analyse the differences between group means and their associated procedures (such as "variation" among and between groups), developed by R.A. Fisher. In its simplest form, ANOVA provides a statistical test of whether or not the means of several groups are equal, and therefore generalises the t-test to more than two groups.
Succesfull Trials:
NFD = 33.76%
NFC = 11.25%
It infers that users are capable to maintain the same error rates, which are relatively low, along by decreasing their competition time.
Finger and wrist fatigue was in extremely low levels.
Finger and wrist fatigue was in extremely low levels.
Finger and wrist fatigue was in extremely low levels.
Fat Finger is capable to become both a sufficient alternative to the current gestured based approaches and also a much intuitive way to perform operations that are infeasible at the moment.
However, the assimilation degree Fat Finger will encounter depends highly on the level of satisfaction, user experience, pleasure and throughput it will incorporate. In the end, we must accept the reality that the future of Fat Finger will depend largely on market trends, which are mainly determined by human beings, their needs and desires.
Grasp zoom: sensor FSR attached on the back of the device.
Pressure text & Clarkson: map multiple taps to different types of pressure
Brewster et al. They map soft presses to lower-case letters, and hard presses to upper-case, trying to boost mixed-case text typing.
Lee et al : It was one of the first three dimensional approaches in interacting with a tablet device.
also gives measure of degree of contact for each finger. enable multi touch using grid. Expand HCI vocabulary.
ShapeTouch: Pressing, friction, dragging, rotating, anchored movement
Fat Thumb: Driving example: Map App: contract size was controlling zooming or dragging mode.
ThumbRock: It can be used as a supplement to tapping, allowing editing or zooming depending on the application.