SCENE

Context Awareness System and Service

1

SCENE

2011.04.27JS Lee

UbiPhone:Human-Centered Ubiquitous Phone System


2

Contents

RESULTS

EXPERIMENTS

IMPLEMENTATION

SOCIAL NETWORK MODEL

CONTEXT DIGEST MODEL

ARCHITECTURE

UBIPHONE FEATURES UBICALL, ANYCALL, EMERGENCY CONTACT SERVICE

INTRODUCTION


3

Introduction

A human-centered personal assistant service must discover how best to reach a contact at any given moment, rather than simply asking users to choose from a list of phone numbers.

It studies smart phones’ potential to become intelligent personal digital assistants.

UbiPhone automatically connects using the most appropriate phone system based on current context information, such as caller and contact’s location, presence status, network status, available phone systems, calendars, and social relationships.


4

UbiPhone Features

Ubiphone adopts context and social models to develop intelligent human-centered services.

Ubiphone’s special features are

UbiCall, AnyCall, and an emergency contact service.


5

UbiPhone Features - UbiCall

UbiCall provides a novel human-centered phone call service. Users who want to call a friend from their contact list simply click on their friend’s name.

UbiCall automatically selects the most appropriate phone number (home, office, or cell phone, VoIP, or IM) and network interface to use, eithercellular (UMTS/HSDPA or GPRS/EDGE/GSM) or wireless (WiFi, Bluetooth, or WiMax), based on context information such as reachability, call rate, and user preference.

Additionally, the service provides intelligent feedback when contactsaren’t available, such as when to call back according to their calendars, or whether to re-dial automatically when contacts become available.


6

UbiPhone Features - AnyCall

AnyCall is similar to the anycast service in IP networks, enabling users to call the most appropriate friend within a group.

Consider an emergency scenario in which a user must call his family to inform them about an accident. In this situation, anyone from the caller’s family list can help him. AnyCall will connect the caller to the first available family member on the list with a single click on the family group’s name.

AnyCall also considers other context based on the caller’s configuration. For example, the system could call the available friend with the cheapest call rate.


7

UbiPhone Features – An Emergency Contact Service

In emergency cases, an emergency contact service could connect the caller to a third party from the contact’s phone number list, or to a phone at a location where the contact can be reached immediately.

The service uses social, calendar, and location information to determine the most appropriate third party.


8

Architecture


9

Context Digest Model

The context digest model adopts an ontology as the underlying technique to manage context.

The context digest model acquires context using a context-acquisition module from various agents and smart devices.

The context-management server classifies, transforms, and stores this data, which it classifies as static, dynamic, statistical, or social.

The static context, such as users’ profiles and preferences, rarely, if ever, changes. The dynamic context, such as users’ presence status and locations, updates the context-management server frequently, and responds to the smart phone immediately. The statistical context, such as phone call logs, location logs and event logs, is derived from statistics of historic logs. The social context provides a social relationship model generated from social activities between two individuals, such as phone calls, emails, or instant messages.


10

Social Network Model


11

Implemetation

It comprised several servers and four users using Dopod PDA phonesequipped with WiFi, GSM, GPS, and RFID modules. We installed RFID readers in rooms to collect each user’s location log.

The users (Septhiorth, Hushpuppy, Lunsrot, and Zqq) were members of the E-learn 2.0 project. Each user also joined one or more research groups, including UbiCom, Web 2.0, and e-learning groups.


12

Implementation


13

Experiments

We performed the experiment for five weeks and derived log statistics from each hour on a weekly cycle. We collected and stored users’ location logs and phone call logs in the context management server during the experimental period. After the fourth week, we use the data from the fifth week to evaluate the emergency contact service’s accuracy. Specifically, weselected a random user, Masa, to be the unreachable target contact. Masa’s contacts, location log, and phone records are shown elsewhere.

If Masa and the suggested contact were at the same location, then we deemed the trial successful; otherwise, we considered it unsuccessful. In this experiment, we set the weighting parameters as follows: a location = 0.65, a relation = 0.35, Bw = 0.3.

Thus, we considered the location context as more important thanthe social context.


14

Results


15

Results

The emergency contact service’s overall accuracy was 77.38 percent for this experiment.

In most cases, Masa was performing a routine activity, and so the trial wassuccessful.

The emergency contact service generally failed to recommend the appropriate third party when Masa encountered a special event or activity which he didn’t place in his calendar, such as donating blood or a special party. Intuitively, the accuracy of the emergency contact depends strongly on whether the activity’s location is predictable and whether the acquired context is sufficient to reason out the correct answer.

Finally, the emergency contact service could consider suggesting more than one candidate to contact if the calculated highest possibility of all contacts is less than a threshold.


16

Q & A Session

Feel free to ask any question

SCENE

Documents

static context

appropriate phone system

current context information

contextacquisition module

contextmanagement server

emergency contact service

contact list

cell phone