A Standards-Based Generic Approach for Complex
Multimedia Management
Anna Carreras1, Ruben Tous1, Eva Rodríguez1, Jaime Delgado1, Giovanni
Cordara2, Gianluca Francini2, Diego Gibellino2
1
Universitat Politècnica de Catalunya, Departament d’Arquitectura de Computadors,
Campus Nord, Mòdul D6, Jordi Girona 1-3, E-08034 Barcelona, Spain
{annac, rtous, evar, jaime.delgado}@ac.upc.edu
2
Telecom Italia Lab, Via Reiss Romoli, 274 - 10148 Torino, Italy
{giovanni.cordara, gianluca.francini, diego.gibellino}@telecomitalia.it
Abstract. This paper presents a standards-based architecture for a complex and
generic distributed multimedia scenario, which combines content search and
retrieval, DRM, and context-based content adaptation together. It is an
innovative and totally generic approach trying to narrow the semantic-gap by
integrating a flexible language for multimedia search based on MPEG Query
Format (MPQF) standard with the application of video analysis algorithms for
the automatic extraction of low-level features and with the use of contextual
information.
Keywords: Context-based adaptation, Digital Rights Management-Rights
Expression Language, Features extraction, MPEG Query Format.
1 Introduction
A substantial amount of work has been done in the area of Universal Multimedia
Access (UMA) and the most recent works focus on the maximisation of user
experience. Nevertheless, these approaches are usually application-specific and it is
easy to identify serious limitations in terms of interoperability and extensibility. The
majority of the research activities reported in this area focus mainly on content
adaptation [1], where the use of contextual information and metadata is essential to
achieve efficient and useful adaptations that enrich the user experience. Furthermore,
the rising tide of available content has created the need of new tools for guiding users
to be able to search and find what is of their interest. In both (content adaptation and
search) applications, similar problems need to be addressed:
• The lack of context and metadata textual descriptors.
• The semantic gap, i.e. the gap between the low-level features (LLFs) that can
be automatically extracted from digital contents and the interpretation that a
user would have of the same content.
• The use of standards.
The affordability of consumer electronic devices such as MP3 and recordable
players, digital cameras, etc., allow users to become content producers as well as
consumers. This evolution creates new and interesting challenges: the
abovementioned issues are tightly related to the industry exploitation of new search
and retrieval solutions, able to address the specific requirements of the emerging
social networks featuring tons of user generated content (UGC). The combination of
techniques taking advantage of automatically extracted features, textual metadata and
context information can represent the key for the success of such services, offering
simple and seamless management of personal digital media repositories in the home
network or on the big Internet, as well as Premium content catalogue browsing and
search features.
Furthermore, a multipurpose framework dealing with heterogeneous contents needs
to take into account the enforcement of Digital Rights Management (DRM)
technologies during content access and consumption. Such a feature represents a key
issue for business models design, ensuring a transparent and correct usage of the
content throughout each stage of the value chain (from the content creator, through
the service provider, and to the end user).
Following the background information about the different lines of research
activities integrated in the theme of the proposed work (Section 2), this paper will
address all the identified challenges by presenting a standards-based architecture for a
complex and generic distributed multimedia scenario, which combines search and
retrieval, DRM and context-based content adaptation (Section 3). Finally, before the
conclusions and the future work, an application scenario based on Social Networks is
described in order to better evaluate our proposal (Section 4).
2 Background
2.1 Multimedia Search and Retrieval
In this section we will first analyse the requirements of today’s multimedia search and
retrieval services, and present the MPEG Query Format (MPQF) as the most suitable
solution as it satisfies those requirements. Finally, we will also refer to different
search and retrievals algorithms based on video processing techniques.
Unified Querying Languages and Interfaces. The first thing to take into account
when defining a search and retrieval service is that user information needs can be
expressed in many different ways. On the one hand, when search preferences can be
expressed in terms of precise conditions as those in a relational algebra expression,
clearly determining which objects of collection to select, it is known as data retrieval
(DR). In this case, a single erroneous object among a thousand retrieved objects
means a total failure. In the context of multimedia search and retrieval, DR refers to
queries expressed in terms of metadata and also in terms of low-level features. On the
other hand, there are user information needs which cannot be easily formalized.
Information retrieval (IR) aims to retrieve information which might be relevant to the
user, given a query written from the user’s point of view. In the context of multimedia
search and retrieval, IR refers to text keywords and query-by-example (QBE) for
instance.
Querying today’s digital contents can imply the combination of data retrieval-like
conditions referred to a well-defined data model and also information retrieval-like
conditions.
Many modern multimedia databases (MMDBs) and various providers of
multimedia search and retrieval services already offer advanced indexing and retrieval
techniques for multimedia contents. However, their databases and service interfaces
are proprietary, and therefore the solutions differ and do not interoperate.
Our proposed search and retrieval service is based on MPEG Query Format
(MPQF) standard in order to guarantee the interoperability needed to ease the access
to repositories by users and applications, and to allow the deployment of distributed
search and aggregation services.
MPEG Query Format Overview. The MPEG Query Format (MPQF) is Part 12 of
ISO/IEC 15938-12, “Information Technology - Multimedia Content Description
Interface”, better known as MPEG-7. The standardization process in this area started
in 2006, and MPQF became an ISO/IEC final standard after the 85th MPEG meeting
in July 2008.
MPQF is an XML-based query language that defines the format of queries and
replies to be interchanged between clients and servers in a distributed multimedia
information search-and-retrieval context. The two main benefits of standardising such
kind of a language are 1) interoperability between parties (e.g., content providers,
aggregators and user agents) and 2) platform independence (developers can write their
applications involving multimedia queries independently of the database used, which
fosters software reusability and maintainability). The major advantage of having
MPEG rather than industry forums leading this initiative is that MPEG specifies
international, open standards targeting all possible application domains, which are not
conditioned by partial interests or restrictions.
Multimedia Search and Retrieval Techniques. Video data can be indexed based on
its audiovisual content (such as colour, speech, motion, shape, and intensity), and
semantic content in the form of text annotations. Because machine understanding of
the video data is still an unsolved research problem, text annotations are often used to
describe the content of video data according to the annotator’s understanding and the
purpose of that video data.
As far as indexing and retrieval techniques for the visual content are concerned,
content-based solutions propose a set of methods based on low-level features such, for
example, colours and textures. Several frameworks dealing with the automatic
extraction of low level features have been proposed [3]; their main disadvantage,
however, relies in the impossibility for such systems to process complex queries to
express high level semantic concepts, like, for example “find a video of my sister on a
beach at the sunset”. Some recent technologies allow the indexing of content based on
high level concepts through specific algorithms, but the categorization is limited to
few concepts due to the implicit constraints imposed by those algorithms. In
TRECVID 2007, the search task consisted of finding shots in a test collection
satisfying queries expressed by topics – a kind of complex high level features.
Examples of such topics are: “waterfront with water and buildings” and “street protest
or parade”. This TRECVID contest regarded 24 topics. The recall and precision
values are still quite modest, as compared to other information retrieval scenarios.
Some opportunities to improve those systems can be offered by the combination of
signal and symbolic characterizations in order to diminish the semantic gap and
support more general queries: this approach allows to take into account low level and
high level concepts and to enable different query paradigms (search by similarity,
search by analogy, etc.).
2.2 Concepts and Models for Context and Metadata
Even if the study of metadata and context has been carried out for many decades,
nowadays there is still some confusion on defining and modelling context and
metadata.
For example, when dealing with Information Services, the dynamic behaviour of
some metadata descriptors is sometimes interpreted as context, as in [4]. In other
works usually focused on mobile applications, such as [5], context metadata is
defined as “information that describes the context in which a certain content item was
created”. And finally, advanced works trying to integrate context and content like [6]
decided that “the term context refers to whatever is common among a set of
elements”.
Context and metadata are clearly associated to knowledge; “meta-data” is
information about data, but what is context doesn’t seem to be so clear. We agree with
the probably most generic definition found in the literature provided by A. Dey [7]:
"Context is any information that can be used to characterize the situation of an
entity. An entity is a person, place, or object that is considered relevant to the
interaction between a user and an application, including the user and application
themselves."
Furthermore, while metadata is supported by really mature standardised schemas
such as the one defined by MPEG-7, there is a clear need of defining a common
schema of contextual information for generic multimedia scenarios. If not, it will be
difficult to take the maximum advantage of it use because the advanced works in the
area (such as the previously identified) won’t be able to be extended and interoperable
with the incoming complex multimedia scenarios of the future.
2.3 Context-Aware Content Adaptation
Context-aware content adaptation has become an important line of research, however
it has always lacked of standardised models to represent and manage contextual
information. Three main initiatives should be identified: the CC/PP (Composite
Capability/Preference Profile) created by the W3C which defines an RDF-based
framework for describing device capabilities and user preferences, the UAProf (User
Agent Profile) of the Mobile Alliance Forum which provides an open vocabulary for
WAP (Wireless Access Protocol) clients to communicate their capabilities to servers,
and the Usage Environment Description (UED) tool included in MPEG-21 Digital
Item Adaptation (DIA) which consists of a complete set of context descriptors for a
multimedia adaptation scenario. It includes user information, network and terminal
capabilities as well as natural environment descriptors.
The first two are limited to specific applications, and represent a small subset of
contextual information only. Without doubt, the most complete initiative trying to
identify and represent the context for generic multimedia applications has been
carried out by the MPEG community by means of the 7th part of its MPEG-21
standard. It is called MPEG-21 DIA (Digital Item Adaptation) and includes all kind of
descriptors to facilitate context-based content adaptation.
2.4 Digital Rights Management (DRM) Initiatives
A DRM system provides intellectual property protection so that only authorized users
can access and use protected digital assets according to the rights expressions, which
govern these assets.
Nowadays, there are several commercial initiatives that specify a complete DRM
system. Moreover, there are standard initiatives that specify the elements that form a
part of a DRM system and the relationships between them. Among the standard
initiatives, the most relevant is the MPEG-21 standard that defines a framework for
dealing with different aspects of multimedia information management. This standard
has normatively specified the different elements and formats needed to support the
multimedia delivery chain. In the different parts of this standard, these elements are
standardised by defining the syntax and semantics of their characteristics, such as the
interfaces to these elements.
In a DRM system rights expressions are defined to be the terms that govern the use
of the digital assets. They are presented to the different actors of the digital value
chain in the form of licenses expressed according to a Rights Expression Language
(REL). RELs specify the syntax and semantics of a language that will be used to
express the permissions and restrictions of use of a digital content. Licenses created
according to a specific REL are associated to digital assets and can be interpreted and
enforced by a DRM system.
Adaptation Authorisation. Adaptation operations should only be performed if they
do not violate any condition expressed in the licenses. MPEG-21 DIA specifies
description formats for permissions and conditions for multimedia conversions that
are useful to determine which changes (adaptations) are permitted on content in view
and under what kind of conditions.
3 Architecture Description
The proposed search and retrieval architecture for a complex and generic distributed
multimedia scenario is depicted in Fig. 1. The core of the architecture is a multimedia
content search module based on MPQF, which is elaborated in Section 3.1.
Furthermore, the content retrieval service is augmented with a context-based content
adaptation service based on MPEG-21 DIA and a DRM service based on MPEG-21
REL, which are discussed in detail in Section 3.2 and Section 3.3 respectively.
Multimedia Content Provider #1
Adaptation
Engine
Context
Authorisation
Engine Content
Database
MPQF MPQF
Client Query
Engine
Service
Provider
MPQF
Multimedia Content Provider #2
Fig. 1. Proposed search and retrieval architecture for a complex and generic distributed
multimedia scenario.
The proposed architecture is flexible and extensible not only because standards
have been used, but also because of its modular structure. It allows many different
types of multimedia content providers ranging from a simple metadata repository to
an advanced provider such as Multimedia Content Provider #1 in Fig. 1. Furthermore,
the application of video analysis algorithms for the automatic extraction of low level
features could also be integrated in order to address the lack of metadata textual
descriptors.
The users can transparently use the DRM engine and the adaptation engine to
enrich the content retrieved from the content databases and the query engine.
Furthermore these services can also be called by the user application, directly. Finally,
the management of contextual information based on profiles (presented in Section
3.2) enriches content adaptation services, as well as content search and retrieval,
whenever necessary; and therefore, will also be addressed. .
3.1 Multimedia Content Search and Retrieval Service
MPEG Query Format-Required search functionalities amongst the modules in the
proposed architecture vary depending on their roles. On the one hand, service
providers (e.g. content aggregators) need collecting metadata descriptions from
content providers, and this is usually performed through a harvesting mechanism.
Metadata harvesting consists on collecting the metadata descriptions of digital items
(usually in XML format) from a set of digital content providers and storing them in a
central server. Metadata is lighter than content, and therefore, it is feasible to store the
necessary amount of metadata with the service provider, so that real-time access to
information about distributed digital content becomes possible without the burden of
performing a parallel real-time querying on the underlying target content databases.
The search functionalities required for harvesting are very simple, because
“harvesters” usually request information on updated records using a datestamp range.
On the other hand, content “retailers”, which include service providers and also
some content providers (generally medium or large scale providers), should be able to
deploy value-added services offering fine-grained access to digital items, and
advanced search and retrieval capabilities.
We have chosen the MPEG Query Format as the interface between parties in either
of the two different situations described below. Although there exist mechanisms for
metadata harvesting (e.g., Open Archives Initiative), MPQF can offer not only a
similar functionality, but also a broad range of advanced multimedia search and
retrieval capabilities. One of the key features of MPQF is that it is designed for
expressing queries combining the expressive style of IR systems with the expressive
style of XML DR systems (e.g., XQuery), embracing a broad range of ways of
expressing user information needs. Regarding IR-like criteria, MPQF offers a broad
range of possibilities that include, but are not limited to query-by-example-
description, query-by-keywords, query-by-example-media, query-by-feature-range,
query-by-spatial-relationships, query-by-temporal-relationships and query-by-
relevance-feedback. Regarding DR-like criteria, MPQF offers its own XML query
algebra for expressing conditions over the multimedia related XML metadata (e.g.,
Dublin Core, MPEG-7 or any other XML-based metadata format) while also offering
the possibility to embed XQuery expressions (see Fig. 2).
XML query algebra
(metadata-neutral)
DR-like criteria
Embedded XQuery expressions
(metadata-neutral)
MPEG Query Format
QueryByFreeText
IR-like criteria
QueryByDescription
QueryByMedia
SpatialQuery
TemporalQuery
Fig. 2. MPEG Query Format Outline
A valid MPQF document (according to the MPQF XML schema) always includes
the Mpeg-Query element as the root element. Below the root element, an MPQF
document includes the Input element or the Output element, depending on the fact
that if the document is a client query or a server reply (there is only one schema and
one root element). The part of the language describing the contents of the Input
element is usually named as the Input Query Format (IQF), which it mainly allows
specifying the search condition tree (Fig. 3) and also the structure and desired
contents of the server output. The part of the language describing the Output element
is usually named as the Output Query Format (OQF), and it specifies what the valid
outputs are from the server to the client. IQF and OQF are used to facilitate
understanding only, but do not have representation in the schema.
let $a := node()//n:CreationCoordinates R0lGODlhFwAOAKI
/n:Date/n:TimePoint/ 8ALAAAAAAXAA4AAANCCLrM1j
JtcdMcd3mGEJ5jRpinSG5s26mw
return ($a > 2007-05-01T00:00:00) AND
($a < 2007-05-30T23:59:59) tGMXiAyibA8wEGBKDQUjgeKcD
dM63VxiyQAADs=
Barcelona
QueryByFreeText QueryByXQuery QueryByExample
AND
Fig. 3. Example Condition Tree
Multimedia Content Search and Retrieval Algorithms. A search and retrieval
service able to respond to generic queries expressed with MPQF needs to be
integrated with effective search and retrieval algorithms.
As stated above, technology that has proven to guarantee good performances in
different use cases is the analysis of textual metadata (keywords, textual descriptions,
plots, actors, user comments, etc.): there are several standards (like MPEG-7)
describing information related to multimedia contents in a textual and interoperable
format. In a real scenario, however, video content is not always accompanied by other
corresponding information. There is the need, therefore, to provide innovative ways to
allow users to search for content exploiting all the available information. A solution
can be identified in the automatic analysis of visual information and MPEG-7
represents a standard way to describe a set of low level features in an interoperable
XML format. Nevertheless, one has to get over the abovementioned semantic gap. A
way to address the problem and improve the overall performances, still maintaining
the standard compatibility is to analyse MPEG-7 descriptors related to low level
features jointly with textual metadata, whenever available.
The technical work conducted in this activity can be described as a sequence of
different operations:
• Automatic extraction of low level features: for each video, a set of MPEG-7
descriptors is extracted; the obtained low level features are then processed in
order to extract temporal and spatial features related to the whole content. The
latter are represented by the MPEG-7 descriptors themselves, providing
information about visual aspects. Each frame is associated to an element of a
codebook made clustering the MPEG-7 descriptors extracted from a training
set of videos, using the Generalized Lloyd Algorithm (GLA). A probability
distribution related to the whole video is then computed. This distribution
describes the visual aspects of the video.
• Starting from the analysis of MPEG-7 descriptors it is also possible to obtain
information about the temporal evolution of the videos, revealing, therefore,
aspects of storytelling style. After having compressed the total amount of data
through the Singular Value Decomposition (SVD), we have chosen to develop
new low level features that reflect the complexity of the temporal evolution of
the principal components extracted by the data set: the spectral flatness and the
fractal dimension. This is because the amount of change among frames can be
inferred directly from the variation of the first few coefficients.
• The visual and temporal features undergo a fusion process to compute the
overall similarity among contents;
• Analysis of textual information: Latent Semantic Indexing (LSI) technique
has been used, a vector space technique that exploit co-occurrences between
terms. Using LSI it is possible to discover similarities between texts even if
they share few or no words;
• Construction of searchable indexes: The data extracted with textual and
visual analysis are used jointly for creating tables of distances between
contents in the repository. Such tables can be used in real time to provide
answers to different kind of queries.
For further information about these algorithms one can refer to [8].
Once the searchable indexes are constructed, we can consider that this Content
Provider (CP) includes not only the Content database (Fig. 1), but also these
processing techniques and the associated database with the similarity measures
between contents. When contents lack of metadata, or even for a more accurate result
based on a user’s Relevance Feedback (RFB), or combined with metadata,
QueryByExample (QBE) is an interesting retrieval approach that needs to be
addressed.
We already stated, at the beginning of this section, that MPQF is capable of
expressing different types of queries for IR systems, as for example, query-by-
example-media In Fig. 4 and Fig.5 we can see an example of the InputQuery and the
OutputQuery, respectively, that could be used between the Service Provider (SP) and
the CP previously identified. On the one hand, the request includes the sample of
content that is used to express user’s interest and the description of the desired output
(number of results, etc.). On the other hand, the response includes the list of the “most
similar” items that have been retrieved.
Moreover, we could consider these messages as the ones exchanged between the
user and the SP in a more specific application scenario.
Fig. 4. Example of an MPQF QueryByMedia Input Query
Fig. 5. Example of an MPQF QueryByMedia Output Query
3.2 Context-based Content Adaptation Service
In a complex multimedia scenario, many different adaptation operations may be
performed on contents. As mentioned in Section 2, MPEG-21 DIA is a complete
standard that specifies the syntax and semantics of tools that assist in the adaptation of
multimedia content. It is used to satisfy transmission, storage and consumption
constraints as well as Quality of Service management. The proposed context-based
adaptation service for search and retrieval application is based on MPEG-21 DIA; a
more detailed specification of a possible modular architecture of this service can be
found in [9].
Due to the inherent complexity of the standards, their practicality in video search
and retrieval domain is slim. In order to address this issue, a number of context
profiles are defined. These profiles include User Profile, Network Profile, Terminal
Profile, and Natural Environment Profile. They contain all the associated descriptors
of MPEG-21 UED, and thus, cover a complete set of contextual information. A
detailed description of them can be found in [9]. Furthermore, they could be extended
to new types of context that could be identified by new sensors.
The use of profiles eases the introduction of standards while providing more
flexibility and scalability to any architecture. In our proposal, their use would
definitely enrich the search and retrieval service while guaranteeing interoperability.
Not only the user can identify the content he/she wants, but also will receive it in the
most optimized way thanks to the Context-based Content Adaptation Service
3.3 Digital Rights Management Service
The DRM Management service will ensure that multimedia copyrighted content is
used according to the terms stated by content creators or rights holders. This service
will inform to the user the operations that he/she can perform with the videos found
by the audiovisual search and retrieval service. It provides functionalities to obtain the
licenses governing a digital resource, in this case a video, and provides information
content usage information, according to licenses governing the selected video, to the
user. Then, the user will select the operation he/she wants to perform, and if necessary
will purchase the appropriate license.
This service provides two operations: The first one obtains the licenses associated
to the video selected by the user and the second one determines the user’s permissions
and constraints of content usage. Next are described the two operations in detail:
• getLicenses: It receives as parameters a video label and returns an XML file
containing the set of MPEG-21 REL licenses governing the image. These
licenses will specify the rights that the user can exercise and the conditions that
he previously has to fulfil. Moreover, it also returns the rights that the use
could exercise if he previously purchases the appropriate license.
• verifiyRights: It receives as parameters the user’s licenses governing the
video and an XML file containing information about the usage that the user
has previously done with this concrete video, and determines if the user can
exercise the requested operation. This operation implements a license
verification algorithm, based on the MPEG-21 REL Authorization Model [x],
which verifies if an entity was authorized to perform the requested operation
over a video.
Adaptation Authorization: In order to govern content adaptations for protected
contents, licenses integrating MPEG-21 REL and MPEG-21 DIA should also be used.
The main reason for integrating both standards is that, due to the increasing
complexity of adaptations, we require more detailed descriptions about content
adaptation in order to govern them. A detailed work on the adaptation authorisation
can be found in [10].
4 Social Networks application scenario
Recently, online social networking sites are experimenting a dramatic growth in
their use. Users of these sites form social networks and share contents (photos, videos,
etc.) and personal contacts. In certain occasions users can wish to protect their
personal information and the contents they share for privacy issues. Online social
networking sites can be accessed from a broad diversity of devices (PDAs, mobile
phones, PCs, laptops, etc.) and in different network conditions (fixed, mobile, local
area, wide area, etc), then an efficient adaptation of the contents is required. Due to
the huge amount and diversity of contents shared by users, online social networking
sites also require efficient search and retrieval solutions. Furthermore, these solutions
also can take advantage of automatically extracted features and textual metadata.
In online social networking sites, users will benefit of the proposed solution, since
service providers can collect metadata descriptions from content providers and
content retailers should be able to deploy value-added services offering fine-grained
access to digital items, and advanced search and retrieval capabilities. Moreover,
access to these sites can be done from a broad range of consumer electronic devices,
since the framework provides efficient and useful adaptations. Finally, users will be
able to protect their personal information, contacts as well as the contents they
provide to other users.
5 Conclusions and Future Work
This paper has presented a standards-based generic approach for complex multimedia
management.
First of all, several weaknesses when dealing with context and metadata for
multimedia management have been identified. The novelty of our proposed solutions
comes from the fact that these problems are addressed in the most generic way, as the
authors consider it is the best approach to exploit the maximum potential of both
types of descriptors.
On the one hand, a flexible and extensible way of representing context is used to
enrich content adaptation, content search, and Digital Rights Management. On the
other hand, the use of a flexible language for multimedia search and retrieval based on
MPEG Query Format is the key point trying to narrow the semantic gap, as it gives all
the required functionalities. Furthermore, the lack of metadata textual descriptors has
been addressed by integrating video analysis algorithms for the automatic extraction
of low-level features with this flexible language for multimedia search and retrieval.
The use of standards, such as MPEG-21 DIA, MPEG-21 REL, MPEG-7 and
MPQF, is also mandatory to guarantee interoperability with similar systems.
We will continue working on the instantiation of the approach presented in this
paper in an ongoing project, the XAC2 project (sequel of XAC, Xarxa IP Audiovisual
de Catalunya, Audiovisual IP Network of Catalonia); XAC2 is a network for digital
assets interchange among TV channels and content producers. It is worth mention that
it is expected that from this work it will emerge the first known implementation of an
MPEG Query Format processor. Currently, parts of the ongoing implementation are
being contributed to the MPEG standardisation process in the form of Reference
Software modules.
Acknowledgments. This work has been jointly supported by the European
Commission IST FP6 programme (VISNET II Network of Excellence, IST-
2005.2.41.5) and the Spanish government (DRM-MM project, TSI 2005-05277).
References
1. Wang, J.-G. Kim, S.-F. Chang, and H.-M. Kim, “Utility-Based Video Adaptation for
Universal Multimedia Access (UMA) and Content-Based Utility Function Prediction for
Real-Time Video Transcoding,” IEEE Trans. Multimedia, vol. 9, no. 2, pp. 213-220,
February 2007.
2. ISO/IEC/SC29/WG11/N9341. “ISO/IEC 15938-12 FCD MPEG Query Format”, October
2007.
3. L. Xu, L. and Li, Y. 2003. Video classification using spatial-temporal features and PCA.
In Proceedings of the 2003 international Conference on Multimedia and Expo - Volume 3
(ICME '03) - Volume 03 (July 06 - 09, 2003). ICME. IEEE Computer Society,
Washington, DC, 485-488.
4. A. Sorvari, J. Jalkanen, R. Jokela, A. Black, K. Koli, M. Moberg and T. Keinonen,
“Usability issues in utilizing context metadata in content management of mobile devices”,
in Proc. of the third Nordic conference on Human-computer interaction (NordiCHI’04),
Tampere, Finland, 2004.
5. M. S. Aktas, G. C. Fox and M. Pierce, “Managing Dynamic Metadata as Context”, in
Proc. Of the 2005 Istanbul Internacional Computational Science and Engineering
Conference (ICCSE2005), Istanbul,,Turkey, 27-30 June 2005.
6. M. Wallace, G. Akrivas, Ph. Mylonas, Y. Avrithis and S. Kollias, “Using Context and
Fuzzy Relations to Interpret Multimedia Content”, in Proc. of the Third International
Workshop on Content-Based Multimedia Indexing (CBMI), Rennes, France, Sep. 2003.
7. A. K. Dey, “Providing Architectural Support for Building Context-Aware Applications,”
Ph.D. Thesis, College of Computing, Georgia Institute of Technology, Atlanta, Georgia,
2000.
8. IST-1-038398 - Networked Audiovisual Media Technologies - VISNET II, “Deliverable
D2.2.5: First set of developments and evaluation for search systems for distributed and
large audiovisual databases”. November 2007.
9. M. T. Andrade, H. Kodikara Arachchi, S. Nasir, S. Dogan, H. Uzuner, A. M. Kondoz, J.
Delgado, E. Rodríguez, A. Carreras, T. Masterton, and R. Craddock, “Using context to
assist the adaptation of protected multimedia content in virtual collaboration applications”,
in Proc. 3rd IEEE Int. Conf. on Collaborative Computing: Networking, Applications and
Worksharing (CollaborateCom 2007), New York, USA, 12-15 Nov. 2007.
10. A. Carreras and J. Delgado, “A new type of contextual information based on the
adaptation authorisation”, accepted to the 9th Int. Workshop on Image Analysis for
Multimedia Interactive Services (WIAMIS 2008), Klagenfurt, Austria, 7-9 May 2008.