the international symposium on wearable computers mark the dates... submissione due: april 7 2013...
TRANSCRIPT
The International Symposium on Wearable
Computers
Mark the dates...Submissione due: April 7 2013 (papers, notes, or posters)Conference: September 9-12, 2013
...and follow us, like us, plus us...www.iswc.net
`
The 12th International Conference on Mobile and Ubiquitous Multimedia
Mark the dates...Submission due : Aug 17 (Short and Full Paper), Oct 14 (Posters/Demos)
Conference: Dec 9–12
...and follow us, like us, plus us...www.mum2013.org
Ulf BlankeWearable Computing Lab - ETH Zürich
Dagstuhl, 3-7 Dec 2012
Human Activity Recognition...
What I have been doing
time continuous data
Recognizing composite activities by decomposition into isolated activity events• Wearable motion sensors • Daily routines, maintenance, or construction tasks
http://www.ulfblanke.de/research/activity-recognition.html
Other projectsPlace recognition
http://www.ulfblanke.de/research/localization.html
Other projectsLocation recognitionSleep studies
Other projectsLocation recognitionSleep studies
Improving the kinectImproving Kinect’s depth camera
http://www.ulfblanke.de/research/kinect.html
Main challenges for the activity recognition research (as a process)
1. „Under“-validated vision
2. Missing conceptualization of activity
Real world problem
1. „Under“-validated vision
1. „Under“-validated vision
Do we know the value of activity recognition?– We assume (healthcare, industrial tasks...)– We should validate more with (end!) users
Did we make progress with respect to our vision?
Example: Activities of Daily Living (ADL)
• Definition given• So far some of top level ADL are well recognized
• Often subset of activities selected – Only 3 to 4 out of 6 top-level categories addressed– Different activities across different papers
• ADL aim also at assessing quality of activities performed
Did we solve any real world problem by recognizing ADLs?
State of the art
Proposal of solution
Research Cycle
Clear problem definition
Analysis
1.
2.
Rosy pictureunclear Hard to evaluate
Yes, technical challenges have to be adressedBut: Our real challenge is the vision/application To get beyond a technical artifact To prepare innovation, impact to society identify new topics for research justification for research (for selection of activities)
Excursion: Lean Start up
„not about asking what customers wants but testing the original vision based on what customers do”
„having an original vision and then refining and testing it along the way through tightly controlled experiments“
Some Lean Startup principles
• Driven by a compelling vision, and rigorously testing each element of this vision.
• Vision & motivation is backed up with data
• The Lean Startup applies to all companies that face uncertainty about what customers will want.
• ....
Main challenges for the activity recognition research (as a process)
1. „Under“-validated motivation
2. Missing conceptualization of activity
2. Missing conceptualization of activity
What is an activity?Activity Theory (A. N. Leont’ev 1978)*
(1) Structure of activityJ. Zacks (2001), (Bobick 1997), (Ampft 2007)...
(2) Object orientedness(Philipose 2004; Buettner 2009)...
(3) Internalization and externalization(Craik, 1967) „mental model“
(4) Mediation(Gibson 1977, Norman 1988)
(5) Development
No formal and sufficient explanation for activity in activity recognition research(Recall discussion)
*c.f. (Vygotsky, 1978), (Engestro m, 1987) or (Rubinstein, 1957) �
Discussion
Our frontier is not technology but vision/application!– Look for use cases– Don‘t rely on intuition:
validate with users: test the vision.– Activity Recognition + HCI as a field?
Is a theoretical concept important?– Standardizing effect?– Problem definition easier?