Docstoc

MTEL-Nguyen

Document Sample
MTEL-Nguyen Powered By Docstoc
					Multimedia Information and Mobile-Learning

              Duc Phuong Nguyen
               Martin Guggisberg
                Helmar Burkhart

              University of Basel
                 Switzerland




                                             1
Challenge 1: Content adaptation
Jobs done: 40
Queues : 30
Status: OK
                                                       Short text can be sent via SMS
Elapsed time: 1 hour
Remain time: 2 hours




What about this text?
 Client-side globus-hostname command is not returning a fully qualified domain name (FQN).
 The submission scripts use globus-hostname to send information to the server about how to contact the client.
 The fix is to make globus-hostname return the FQDN.
 There are firewalls and ports (GRAM: 2119, LDAP: 2135) which are not open.
 Check that GLOBUS_TCP_PORT_RANGE is set to 3000 to 3090 on the client
 (or another range acceptable for the server to contact the client on),
 and that all intermediate firewalls are open for this range in both directions.
 GRAM Job submission failed because the executable file permissions do not allow execution".
 Probable cause: You submitted a script, e.g., globus-job-run symphony:2119/jobmanager-loadleveler $PWD/simple.
 ll but simple.ll does not have the execute permission bit set.
 The error msg "Authentication Failed remote certificate signed by unknown CA"
 means the client cannot find the CA for the cert presented by the server.
 Make sure the CA cert is present in the trusted cert dir set with setup-gsi.
 Also, check the time is consistent across the machines.
 The error msg "GRAM Job submission failed because the connection to the server failed (check host and port)
 Make sure that the gatekeeper service is enabled.
 The error msg "GRAM Job submission failed because data transfer to the server failed"
 means the job manager has problems opening some file.
 Make sure $GLOBUS_LOCATION is readable by all the users (mode 755)

 How can we sent this information to mobile users?
                                                                                                                  2
Challenge 2: Content overview

   Difficult to get the most important message first:
„You have 10 voice messages in your mail box…The first message started on
   10.12.2006 at 14h …
…………………………………………………
…………………………………………………
The second message started on 10.12.2006 at 17h …
…………………………………………………

The ninth message started on 10.12.2006 at 17h …
…………………………………………………
The last message is actually what important to me in this example“




                                                                            3
Challenge 3: Integration

Suppose that we have a system that solves

Challenge 1 (content adaptation)
Challenge 2 (content overview)

How can we integrate it into our E-Learning system?




                                                      4
     Solution 1: Content adaptation

                                                             Speech1.wav
                                                Audio
                                 CoMobile

     Long texts                                              Speech2.wav



                                                                   Voice call




• Quite difficult to convert mathematic formulae to speech



                                                                            5
        Solution 2: Content overview

             Speech1.wav
                               Audio                     Text
                                            CoMobile            Keywords extraction
             Speech2.wav                                         (Post processing)



                                                 SMS            1. Lunch
                                                                2. Food
Voice call      Recorded lectures                               3. Project meeting
                                                                4. Concert
                                                                ……………..
                                    List of keyword can be used to
                                    identify the important voice messages

                                                                                     6
CoMobile framework- Uni Basel




- Access learning materials             - Search, notification
- Collaboration (sharing, discussion)   - Self-assessment, voting, quiz


                                                                          7
CoMobile framework- Uni Basel




- Access learning materials             - Search, notification
- Collaboration (sharing, discussion)   - Review, self-assessment, voting, quiz


                                                                                  8
CoMobile-Example

               Voice message
                                            Voice message

                                 CoMobile

Alice
                                                             Bob
        Voice and Text

                                             Text message



                                                            Christian
                         Pengi                                          9
CoMobile-Example

          Voice message
                                     Voice message
            1
                          CoMobile         2



   Voice and Text

   4                                  Text message
                                            3



                                                     10
CoMobile framework-Overview

1.   Content adaptation: long texts are converted into
     speech and then can be accessed by mobile phones

2.   Search on audio data

3.   Platforms and languages independence

4.   Flexible: easy to integrate via CoMobile-services



                                                         11
CoMobile framework-Components (I)

   All components are open source software:
- CoMobile main component
- Plug-in components:
    - Text to Speech: Festival
    - Speech recognition: Sphinx
    - Optical character recognition: GOCR


Plug-in components can be replaced without many modifications.




                                                                 12
CoMobile framework-Components (II)
  Service-oriented: works on document‘s level
(e.g: input=audio files, output=text files)

   CoMobile calls the component‘s public service

   The component invokes CoMobile callback service
    when it finish the task

   Communication between CoMobile and the plug-in
    components are done via HTTP

                                                      13
CoMobile framework- Easy to integrate

   Service-oriented simplifies the intergration process:




PHP-based system with CoMobile   Java-based system with CoMobile
                                                                   14
CoMobile framework-Components (III)

   Festival-Text to Speech: support many languages
    (English, German, French, Spanish, etc.)

   Sphinx-Speech recognition: free acoustic model for
    English (64.000 words; WER = 18%)

   Optical Character Recognition: GOCR, tesseract
    (supported by Google). Recognition rate is still quite
    low for handwritten notes (accuracy rate= appr. 20%
    with empirical tests).

                                                             15
Open issue: Recognition of notes taken by
mobile phone camera
   Handwritten
    notes and the
    low resolution
    of mobile
    phone‘s
    camera
    decrease the
    accuracy of
    OCR systems
    significantly




                                            16
        CoMobile example: Existing Web-based applications can use mobile
        phones to access and submit information by integrating with
        CoMobile services




http://informatik.unibas.ch/courses/multimedia/lecture1.mpg

          Text retrieval, features extraction, vector space retrieval, inverted list, ranking
                                                                                         17
Uni Basel-CoMobile features
   Content adaptation and content overview for mobile
    devices.

   Data submission: send information (text, images,
    audio, video) via SMS, MMS, voice recording.

   SMS notification service (replies, reminds, alerts).

   Access information via various channels: Voice, SMS,
    MMS, WAP, Bluetooth.

                                                           18
Transformation issue

                      Graphic




           Text                    Audio




   Information loss in transformation chain
   Recognition accuracy
   Quality of search: keywords, semantic
                                               19
Conclusion

   Content transformation is an important issues in mobile
    applications.

   Uni Basel-CoMobile: an easy to integrate and flexible
    framework that support collaborative learning with
    mobile devices.

   The CoMobile framework is open source and look
    forward for contributions.


                                                            20
Uni Basel-CoMobile-Demo




        CoMobile demonstration




                                 21
Discussion

   Cost: Users have to pay for SMS, MMS, as well as voice
    call. „Flat rate“ package will be common in the near
    future.
   Storage: 1 minute video = 1 MB (approximately). There
    exists 2GB flash card for mobile phone
   Bandwidth: 14.4 kbps (GSM), 140 kbps (GPRS), 1920
    kbps (UMTS)




                                                         22
SMS gateway - Kannel



                       HTTP request to forward SMS Text



                        RS-232
                     AT Commands
                   AT+CNMI=1,2,0,1,1
SMS Text




                                                     23
     MMS gateway - Mbuni


                           Mbuni
                             HTTP request to forward MMS



                            RS-232
                         AT Commands
                       AT+CNMI=1,2,0,1,1
       MMS
Text, Picture, Audio




                                                           24
PBX Asterisk



                              AGI: Perl, Java, PHP scripts are called




Voice                     PCI socket




         Telephone line

                                                                25

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:5
posted:9/14/2011
language:English
pages:25