Tuesday, May 12, 2009

Pervasive 2009 in Nara

Toshio Iwai gave a fascinating keynote today at Pervasive 2009.
He speaker startet with a very personal introduction about his childhood and how he got motivated to start his electronic arts work: he was inspired by afterimages, animation flipbooks, punchcard harmoniums.

The speaker startet with a very personal introduction about his childhood and how he got motivated to start his electornic arts work. He spent quite some in showing the underlying concept which gave inspiration to his work: afterimages, old-style animation flipbooks, punchcard harmoniums. This was kind of slow in the beginning but really explained the idea of his Teniro-on pretty well:
The idea is that everybody can be a composer in 5 minutes - I'll have to find out sometime myself. Besides that he showed impressive visualization of music. The sound of a piano professional was captured and depicted as flowing light beams emitted from the piano as fireworks depending on speed and pitch of the sounds. A more detailed report here.
Than an interesting paper followed by John Krumm. It suggested to create random, but realistic, road-trips in order to hide personal routes requested from central routing services such as google maps. The paper spend tremendous efforts into determining realistic starting and end-points (e.g. not starting at a bridge), interpolating between waypoints for realistic accleration patterns and even creationg real GPS noise [1]. Obviously, the prove of the realisticness of the generated routes was hard to prove. There was also some discussion whether this approach really holds, e.g. what if you do an 'unusual' trip starting from a bridge, than you will be detected - couldn't you just mix routes of others?
Stephen Intille presented a solution [2] on how to save heating energy. Since programmable thermostats are rarely used (in the US) it is better to switch of heating once people are way and to heat up when they come up - which can be detected by GPS in their phones. He spend quite sometime on commute time statistics and simulating his results. However, he admitted that programmable thermostats will be still better than his rather dull approach: it does not even distinguish between rooms. He argued about persuasive approaches to convince the users to use their thermostats but that was not really shown. It is really questionable whether this lead into the right direction or just supports lazy behaviour of "don't care" and even discourages the usage of thermostats at all as mentioned in the discussion. Btw, nobody talked about remote-controlled themostats from the phone...
Jeffrey Hightower [3] presented to infer user identity from the way of using handling the remote-control. This, definitely, does not solve the world's most sever problems but can help to personalize TV recorder options and could be used to improve TV analysis, e.g. Nielsen branding today only achieves 80% accuracy. A feasibility study of 5 colleagues watching at least 1h revealed significant enough different behaviours - before/during/after button press were used as input from the accelerometer -video capture was used for ground truth. I really like the one slide summary at the start and end of the talk.
Florian Alt [4] tackled the interesting question of which advertisements people would be willing to display on their cars, see also here.
Stacey Kuznetsov [5] presented an interesting paper that prove a successful candidate of showing an idea that did not work. They tried help people to learn new activities by applying haptic feedback with vibration motors: two medium pulse, strong oscillating, single pulse. The applied blending test, auditory test of learning mandarin and visual recognition. The resuls was that low performers improve a little bit using their system...but free recall is even performed worse. Cues were seen as distracting or pulses were too intense. The lessons to aim at reducing cognitive load, adjusting intensity of pulses to individual perceptions and to add semantic meaning to cues were worthwhile even the initial idea of the authors did not work out.
Jeffrey Hightower [6] presented another paper where they investigated privacy concerns with personal sensing in a 4month study of experience reactions to sensing - people wore sensors and were asked about their reaction towards new sensors. Not surprisingly, there were no effect about barometer or accelerometer, big very against audio, because of the need to explain to others you were recording them now (in social and professional contexts: laywer-client, doctor- patient). The conclusions were to design systems with minimal invasive sensors, not rocket science, but a well written paper. In the conference there was some debate about whether the results had been defined by how the questions where asked: e.g. omitting the fact of explaining the mining opportunities of accelermoter, re-transforming of filtered audio.
Finally a paper really enjoyed [7] was on using battery free tags to measure power transitions from transient signals emitted from the powerline. This was a proof of concept paper documenting a great idea. The discussion was a little bit bizarre as revealed culture and technological differences: "But your systems is limited by range, this means it does not work if lines are build into houses' concrete or stone walls?" - "of course not, but which houses are built of stones and concrete"...
Last but not least, Mikko Lehtonen presented our paper [9] on using synchronized secrets, meaning storing consecutive random numbers on RFID tags and infrastructure to protect state-of-the-art RFID tags from cloning. The discussion was heavily focused on the random number generator which we are free to choose in our approach.
In addition to the papers there was also a demo and poster, where we also could present our car claim report system: the mobile phone as an aid in sending the details of a car accident to the insurance company [8].
Btw, Albrecht Schmidt wrote a nice report on Pervasive 2009, go here to see his paper favourites.

Pervasive 2009 just again showed me the same pattern of successful papers that has emergeed throughout the last years: come up with a new interaction technique (e.g. force sensing), display (e.g. car ads), but that's not enough. Conduct a thorough user study which reveals unexpected results (e.g. bend is more difficult than twist), do extensive testing. Of course, sometimes answering whether the underlying phenomenon is actually worth all the methodology effort of evaluating it fades away...

[1] John Krumm, "Realistic Driving Trips For Location Privacy", Seventh International Conference on Pervasive Computing (Pervasive 2009), May 11-14, 2009, Nara, Japan.
[2] M. Gupta, S. S. Intille, and K. Larson, "Adding GPS-control to traditional thermostats: An exploration of potential energy savings and design challenges" Proceedings of the Seventh International Conference on Pervasive Computing, May 11-14, 2009, Nara, Japan.
[3] Keng-hao Chang, Jeffrey Hightower, Branislav Kveton: Inferring Identity Using Accelerometers in Television Remote Control, Proceedings of the Seventh International Conference on Pervasive Computing, May 11-14, 2009, Nara, Japan.
[4] Florian Alt, Christoph Evers, Albrecht Schmidt: Users' View on Context-Sensitive Car Advertisements, Proceedings of the Seventh International Conference on Pervasive Computing, May 11-14, 2009, Nara, Japan.
[5] Stacey Kuznetsov, Anind K. Dey, Scott E. Hudson: The Effectiveness of Haptic Cues as an Assistive Technology for Human Memory, Proceedings of the Seventh International Conference on Pervasive Computing, May 11-14, 2009, Nara, Japan.
[6] Predrag V. Klasnja, Sunny Consolvo, Tanzeem Choudhury, Richard Beckwith, Jeffrey Hightower: Exploring Privacy Concerns about Personal Sensing, Proceedings of the Seventh International Conference on Pervasive Computing, May 11-14, 2009, Nara, Japan.
[7] Shwetak N. Patel, Erich P. Stuntebeck, Thomas Robertson: PL-Tags: Detecting Batteryless Tags through the Power Lines in a Building, Proceedings of the Seventh International Conference on Pervasive Computing, May 11-14, 2009, Nara, Japan.
[8] Baecker, O., Michahelles, F., Bereuter, A., Mollnau, D., Geller, F., & Fleisch, E. : Mobile First Notice of Loss: Web Service-Based Enterprise Integration of the Android Platform, Demo. In Adjunct Proceedings of the Seventh International Conference on Pervasive Computing (pp. 257-260). Vienna, Austria: Austrian Computer Society.
[9] M. Lehtonen, D. Ostojic, A. Ilic, F. Michahelles: Securing RFID systems by detecting tag cloning, In proceedings of H. Tokuda et al. (Eds.): 7th International Conference, Pervasive 2009, Nara, Japan, May 11-14, 2009. LNCS 5538, pp. 291–308.

1 comment:

evision said...