![](bg7.png)
meta-data extracted by media analyzers
(speech2keyword generator). The text data will be
continually processed using text processing and
statistical algorithms in order to better describing
shot, sequences or all media file. For instance, the
video composition allows automatic video
composition by keyword (Figure 7). First a user or
an application performs an HTTP request with a
“keyword”, a user and the name of the rule as
parameters (1). The keyword is processed (2) and a
search request is sent to the database (3). Using
rules, the automatic video editor (4) compose an
XML descriptor which chain different video shots
matching with the keyword. Audio file can be added
in the video composition if the keyword matches
with auto description. The XML descriptor of the
composition is stored in the database (5). An end
user can perform the following function using the
web based video editing application: play the
composition (6 and 7), modify the video
composition (8) and render the composition by
converting the XML descriptor on video format.
5 MODEL AND ALGORITHM
VALIDATIONS
The video editing testbed has been built for studying
and testing automatic video editing based on text
data. To experiment the automatic video editing
algorithms, we will use a video dataset with multiple
semantic concepts in using video. We propose two
methods to experimentally validating models and
algorithms of automatic video editing: Mashup
validation by users and comparison of the mashups
to one or more reference mashups (depending on the
context). For the first methods, the algorithms will
use the user profiles. The user feedbacks analysis
information will be processed and injected in the
algorithms in order to improve future compositions.
For the second, representative users will create
reference mashups in different domains or subjects.
The algorithms will be tested the mashup compared
to reference mashups.
6 CONCLUSIONS AND FUTURE
WORKS
In this paper, we have proposed an approach of
automatic video editing which is derived using
algorithm and rules based on keyword. The text data
is collected in three ways: direct user annotations,
implicit annotation during the video edition and by
extracting keywords from video analysis. For the
future we plan to create more complex models and
algorithms of video composition to allow composing
video from a sentence.
Figure 7: Automatic video editing engine.
REFERENCES
Hua, X.-S., Zhang, H.-J., 2009. Automatic Home Video
Editing, Signals and Communication Technology,
Springer, 353-386.
Hua, X.-S., Zhang, H.-J., 2003. AVE - Automated Home
Video Editing, ACM MM.
Ma, Y. F., Lu, L., Zhang, H. J., Li, M. J., 2002. A User
Attention Model for Video Summarization. ACM MM,
533–542.
Müller Arisona, S., Müller, P., Schubiger-Banz, S.,
Specht, M., 2008. Computer-Assisted Content Editing
Techniques for Live Multimedia Performance, R.
Adams, S. Gibson, and S. Müller Arisona (Eds.):
DAW/IF, CCIS 7, 199–212.
Takemae, Y., Otsuka, K., Yamato, J., 2005. Automatic
Video Editing System Using Stereo-Based Head
Tracking for Multiparty Conversation, CHI 2005,
1817-1820.
Takemae, Y., Otsuka, K., Yamato, J., 2005. Development
of Automatic Video Editing System Based on Stereo-
Based Head Tracking for Multiparty Conversations,
IEEE.
Mudhwuchutyula, C. L., Kankunhalli, M. S., Mulhem, P.,
2004. Content Based Editing of Semantic Video
Metadata, IEEE International Conference on
Multimedia and Expo.
Foote, J., Cooper, M., and Girgensohn, A., 2002. Creating
Music Videos Using Automatic Media Analysis, ACM
MM.
Beaufays, F., Sankar, A., Williams, S., Weintraub, M.,
2003. Learning Name Pronunciations in Automatic
Speech Recognition Systems, Proceedings of the 15th
IEEE International Conference on Tools with
Artificial Intelligence.
Text processing
+
Video editor
+
Algorithms
(rules)
Multmedia
database
End-user Web 2.0
Manual Video
Composer
XML descriptor
Text data
And meta-
data
Data
keyword
function compose($Sentence, $id, $method=“algorithm name")
1
Media streaming
2
3
4
5
6
7
XML2AVI
converter
VideoXML
9
8
http://<server>/cove/api/get/compose/
keyword/<rule>/<user>/<keyword>
Keywords-based Automatic Multimedia Authoring in the Cloud
147