Presentation 11/15: Context-Aware Mobile Computing: Learning Context-Dependent Personal Preferences from a Wearable Sensor Array

[Su Kim]

Authors: Andreas Krause, Asim Smailagic, and Daniel P. Siewiorek,
Title: "Context-Aware Mobile Computing: Learning Context-Dependent
Personal Preferences from a Wearable Sensor Array,"
Published: IEEE Transactions On Mobile Computing, VOL. 5, NO. 2,
February 2006 [7]

[Ayan Banerjee]

Please find my critique in the attachment.

[Su Jin Kim]

Mike's critique

[Su Jin Kim]

Swaroop's critique

[shayok.ch...@asu.edu]

Hi,

Please find the summary of this paper below:

Paper : Context-Aware Mobile Computing: Learning

Context-Dependent Personal Preferences from a Wearable Sensor Array

This paper proposes a novel technique of context aware mobile
computing where a wearable computer alters its behavior based on the
information acquired about its user's state. By exploiting context
information judiciously, the demands of human attention can be
minimized. A two-step approach is described which consists of a
context classifier and a preference learner. The context classifier
identifies contexts and classifies acquired signals. Unsupervised
machine learning techniques are used to define and identify contexts.
The preference learner relates user interactions with the current
context. Bayesian networks are used to learn context dependent user
preferences and provide feedback to the context classifier. Contexts
are merged or splitted based on user preferences and behavior within
the different contexts. The authors have described a case study in the
paper called SenSay which is a context aware cell phone that can adapt
itself to changes in user's state and surroundings. The results
indicate that context aware wearable computers as described in the
paper are suitable for real life applications.

Advantages of the method
1. The application described can be personalized to each user by
learning their individual preferences rather than having a predefined
model. Also, the method does not require external supervision by the
user.
2. The authors have considered multiple forms of context awareness,
in the forms of time, location and activities.
3. A generic Java framework has been described which makes it readily
integrable with other applications.

Drawbacks of the method:
1. The method described is an expensive one - it requires a PDA,
laptop and sensors for each person.
2. The user study is not thorough. Only a particular class of users
having a particular work schedule is taken for the survey.


Regards,

Shayok.

On Nov 20, 1:24 pm, "Su Jin Kim" <sujin...@gmail.com> wrote:
> Swaroop's critique
>

> On Nov 20, 2007 11:25 AM, Su Jin Kim <sujin...@gmail.com> wrote:
>
>
>
> > Mike's critique
>

> > On Nov 19, 2007 11:17 AM, Ayan Banerjee <abane...@asu.edu> wrote:
>
> > > Please find my critique in the attachment.
>

> > > On 11/13/07, Su Kim <sujin...@gmail.com> wrote:
>
> > > > Authors: Andreas Krause, Asim Smailagic, and Daniel P. Siewiorek,
> > > > Title: "Context-Aware Mobile Computing: Learning Context-Dependent
> > > > Personal Preferences from a Wearable Sensor Array,"
> > > > Published: IEEE Transactions On Mobile Computing, VOL. 5, NO. 2,
> > > > February 2006 [7]
>
>
>

> Critique_Swaroop.pdf
> 8KDownload- Hide quoted text -
>
> - Show quoted text -