The document discusses the development and application of empathic glasses that enhance remote collaboration by integrating gaze tracking, emotional recognition, and head-mounted display technology. Key findings indicate that both eye-tracking and remote pointer cues significantly improve task performance and user experience during collaborative tasks. Future work is suggested to enhance emotion recognition and physiological monitoring to further support both explicit and implicit communication in remote settings.

1. EMPATHIC GLASSES:
SHARING REMOTE GAZE AND EMOTION
Mark Billinghurst
mark.billinghurst@unisa.edu.au
February 18th 2016
Christchurch, New Zealand

2. AR Smart Glasses Are Coming
Microsoft Hololens
DAQRI SmartHelmet
Meta SpaceGlasses

3. Ideal for Remote Collaboration
• SmartGlasses are a natural platform for remote collaboration
• Taskspace vs. Talking Head conferencing
Local User
Remote Expert

4. Lessons From Previous Research
• Sharing communication cues improves task performance
• Shared remote view, Remote pointing
• Natural gesture support
• AR can improve further
• Use of HMD, world stabilized AR cues, shared virtual models

5. Communication Model
• Communication mixture of Explicit and Implicit cues
• Explicit
• Conscious cues – speech, gesture, object manipulation, etc
• Implicit
• Unconscious cues – gaze, emotion, face expression, etc

6. Research Questions
• Will supporting both Explicit and Implicit communication
cues improve remote collaboration?
• Can we use technology to provide better communication
than face to face collaboration?

7. Gaze and Video Conferencing
• Gaze tracking
• Implicit communication cue
• Shows intent
• Task space collaboration
• HMD + camera + gaze tracker
• Expected Results
• Gaze cues reduce need for communication
• Allow remote collaborator to respond faster

8. Equipment
• Custom eye-tracker
• Head mounted camera
• Head mounted display

9. Experiment Set Up
• Lego assembly
• Two assembly areas
• Remote expert

11. Experiment Design
• 4 conditions varying eye-tracking/pointer support
• 13 pairs subjects
• Measures
• Performance time
• Likert scale results, Ranking results, User preference

12. Task Performance
• Performance Time (seconds)

13. Ranking
• Median Ranking Values

14. Key Results
• Both the pointer and eye tracking visual cues
helped participants to perform significantly faster
• The pointer cue significantly improved perceived
quality of collaboration and co-presence
• Eye-tracking improved the collaboration quality,
and sense of being focused for the local users, and
enjoyment for the remote users
• The Both condition was ranked as the best in user
experience, while the None condition was worst.

15. Empathic Glasses
• Combine together eye-tracking, display, face expression
• Impicit cues – eye gaze, face expression
++
Pupil Labs Epson BT-200 AffectiveWear

16. AffectiveWear – Emotion Glasses
• Photo sensors to recognize expression
• User calibration
• Machine learning
• Recognizing 8 face expressions
Masai, K., Sugiura, Y., Suzuki, K., Shimamura, S., Kunze, K., Ogata, M., ... & Sugimoto, M. (2015,
September). AffectiveWear: towards recognizing affect in real life. In Proceedings of the 2015 ACM
International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2015 ACM
International Symposium on Wearable Computers (pp. 357-360). ACM.

17. Integrated System
• Local User
• Video camera
• Eye-tracking
• Face expression
• Remote Helper
• Remote pointing

18. System Diagram
• Two monitors on Remote User side
• Scene + Emotion Display

19. Pilot User Evaluation
• Shared block picture building – 5 minutes
• “Make a house”
• 4 Conditions
• V: Video P: Video + remote pointer
• E: Video + expression A: All conditions (V+P+E)
• 5 pairs subjects

20. Measures
• Pictures made
• Likert scale questions (1= disagree, 7 = agree) [No sig. diff.]
• “I felt connected with my partner”
• Ranking of conditions (1=best, 4 = worst) [Sig. diff.]
• “Which condition did you communicate best with your partner in?”

21. Ranking Results
"I ranked the (A) condition best, because I could easily
point to communicate, and when I needed it I could check
the facial expression to make sure I was being understood.”
Q2: Communication Q3: Understanding partner
HMD – Local User Computer – Remote User

22. Behavior Observation
• Video Only
• Remote users talked more
• Described block colour/shape
• Pointing Conditions (A, P)
• Diectic language – “move that over there”
• Expression Conditions (E,A)
• Looked at expression less than 20% of time

23. Lessons Learned
• Pointing really helps in remote collaboration
• Makes remote user feel more connected
• Gaze looks promising
• shows context of what person talking about
• More work needed on emotion/expression cues
• Limitations
• Limited implicit cues
• Two separate displays
• Task was a poor emotional trigger
• AffectiveWear needs improvement

24. Future Work
• Add additional physiological monitoring
• Empatica E-4
• Provide physiological feedback
• Integrated eye-tracking/display

25. Emotion Classification
• Classify emotions from physiological data
• 8 emotional states
• Collaboration with Sensaura
• http://www.sensauratech.com

26. Conclusions
• Wearable systems ideal for task space conferencing
• Remote shared view, remote pointing
• AR can enhance remote collaboration
• Virtual annotations, content sharing
• Additional sensors can be used for additional cues
• Support Explicit + Implicit communication
• Significant areas for future research
• Capturing/sharing physiological cues
• Representing cognitive/emotional state

27. www.empathiccomputing.org
@marknb00
mark.billinghurst@unisa.edu.au