Magic pointing iswc2015

IT UNIVERSITY OF COPENHAGEN

MAGIC pointing for eyewear computers

Shahram Jalaliniya

Diako Mardanbegi

Thomas Pederson

IT UNIVERSITY OF COPENHAGEN

Why gaze-based interaction on eyewear computers?

• Need for touch-less interaction- Interaction on the move- In parallel with real world tasks

• Head & Eye movements as input modalities

• Eyewear device as a platform for gaze-tracker

IT UNIVERSITY OF COPENHAGEN

Challenge of gaze tracking on eyewear computers

Small size of the head-mounted displays & need for more accurate gaze trackers

Eye FOV: 95°

HMD HMD FOV: 14°

(Google Glass)Big screens

HMD

Accuracy of gaze tracker

IT UNIVERSITY OF COPENHAGEN

Target acquisition on GUI

• Point & select• Eye-pointing is fast but:

- Works for only big targets

• Head-pointing is accurate but:- Tiring specially for long distances

+

Target size

Distance fro

m

start p

oint

+

Accuracy of gaze tracking

*

Initial point

Target

IT UNIVERSITY OF COPENHAGEN

Magic (Manual & gaze input cascaded) pointing

• Use gaze as an implicit data to move cursor as close as possible to the target (Zhai,99)

- Liberal approach (fixed distance)- Conservative approach (involve user)

+

+*

Gaze moves the cursor

Close to the target

Manual part of the pointing (Head movement)

IT UNIVERSITY OF COPENHAGEN

Previous Work

Dimensions

Manual part of pointing

Display setup Strategy

MAGIC pointing (Zhai, S., Morimoto, C.,1999)

Mouse Stationary Liberal/Conservative

MAGIC-Touch System (Drewes, H., and Schmidt, A., 2009)

Touch-sensitive mouse

Stationary Liberal

Look & Touch (Stellmach, S., and Dachselt, R.,2012)

Mouse Stationary Conservative

Head assisted gaze pointing(Spakov & et al., 2014)

Head movements Stationary Liberal

Our study Head movements Head-mounted display

Liberal

Previous work

IT UNIVERSITY OF COPENHAGEN

Experiment goal

• Investigating utility of MAGIC pointing for eyewear computers (combination of eye & head movements)

• Comparing user performance in a target acquisition task using:

•Head pointing

•Combination of eye & head (MAGIC) pointing

•Eye pointing

IT UNIVERSITY OF COPENHAGEN

Head pointing

IT UNIVERSITY OF COPENHAGEN

Magic pointing

IT UNIVERSITY OF COPENHAGEN

Experiment Setup

• Participants- 16 participants (mean age=29, 2 females)- all highly skilled computer users (x ̅= 4.62, σ= 0.5)

IT UNIVERSITY OF COPENHAGEN

Apparatus• Head tracker of Google Glass (9DoF)• Home-made monocular wearable gaze tracker (GlassGaze)

with accuracy of 1 degree (about 50 pixel on Google Glass)

IT UNIVERSITY OF COPENHAGEN

Design (2 x 2 x 2)

• Two pointing modalities: - MAGIC & Head pointing

• Two target sizes- Small (30 pixels ≈ 0.6°) & Big (70 pixels 1.4°)

• Two distances- Near (100 pixels ≈ 2°) & Far (280 pixels ≈ 5.6°)

• 15 repetitions- Median of 15 trials is taken

IT UNIVERSITY OF COPENHAGEN

Experiment results: Task completion Time

• Head pointing is faster for the short distance

• MAGIC pointing is faster in longer distance

• Target size effect is observed

• No distance effect is observed in MAGIC pointing for big targets

IT UNIVERSITY OF COPENHAGEN

Experiment results: Error rate (wrong tap)

MAGIC pointing is more accurate for pointing to small

and far targets

IT UNIVERSITY OF COPENHAGEN

Conclusions

• MAGIC pointing makes it possible to select targets smaller than accuracy of the gaze tracker on head-mounted displays

• MAGIC pointing is faster than Head pointing just for long distances

• In MAGIC pointing the distance does not affect the performance for big targets since in this case MAGIC pointing is similar to eye pointing

IT UNIVERSITY OF COPENHAGEN

Outlook

• Investigating different strategies for positioning the cursor to the initial point of the manual part in MAGIC pointing

- Fixed minimum distance- Fixed time

• Investigating MAGIC pointing for UI development

IT UNIVERSITY OF COPENHAGEN