-This paper describes the extensibility mechanism that has been incorporated to the CAIN Adaptation Engine, that provides audiovisual content adaptation based on user preferences, network capabilities and terminal limitations. The integration of new adaptation modules needs no code modifications in the core system, so it does not have to be recompiled for adding or modifying adaptation modules.
The development of both, new access networks providing
multimedia capabilities, and a wide and growing range of
terminals, makes the adaptation of content an important issue
in future multimedia services. Content adaptation is the main
objective of a set of technologies that can be grouped under
the umbrella of UMA (Universal Multimedia Access ) [
First of all, we need a collection of CATs together with the
descriptions of their adaptation capabilities [
For performing an adaptation we need to receive, at least,
the input media content and a usage environment profile. It is
desirable, but unusual, to receive the associated content
description file. Therefore, the required media information
description is usually extracted directly from the content,
using ffmpeg library[
The EM collects all the information extracted from the
required description files (media, usage environment, CAT
capabilities) and launches the DM. The DM selects both the
best CAT and its configuration parameters using a constraints
satisfaction and optimization algorithm [
CAIN considers the integration of new CATs, as well as the update of existing ones. For this purpose, it provides a flexible extensibility mechanism in order to avoid recoding of the core of CAIN as well as recompiling. One of the integration requirements is to make the CAT´s capabilities accessible from the DM. For this, we need to have a complete and lack of ambiguity file, describing what transformations a concrete CAT is able to do and with which input and output formats. Another important requirement for CAT’s integration is the compliance with interface defined for making the CAT accessible by the CAIN core.
The main differences with the previous CAIN
version[
All the files implementing the CAT, or used by it, have to be packaged in a *.jar file. This file has to share name with the CAT it represents.
For integrating a new CAT we need, at least, two files:
x The CAT code in a *.java file, with the same name as the
CAT. This code file should implement the
adapt(…)method, declared as abstract in its parent
class, CATImplementation.java.
x The CAT Capabilities description[
Apart from these mandatory files, it is also be necessary the inclusion of any library used by the CAT. Any library should be packaged together with the mandatory CAT code file. The Adapt(…) method constitutes the common core to every CAT. It is the method which, by a generic interface, performs the different adaptations. It will be called by the EM and it is expected to adapt in a proper fashion, because the DM has chosen it after checking out, in the CAT´s capabilities description file, that it is able to do so. It returns a list containing the paths and formats of the adapted contents. It can return more than a unique output. A CAT, for example, could convert a video into a sequence of images (e.g., keyframes). In this case, the returned structure would contain the paths and characteristics of each of the output images.
The files of each integrated CAT should be packaged in a *.jar file. This file is stored in a CATs’ folder together with the other available CATs. When CAIN execution starts, the CATs’ folder is checked uncompressing each *.jar file in an execution space accessible by the EM and the DM.
The proposed extensibility mechanism allows implementation-agnostic integration of new CATs in order to incorporate new functionalities without prior knowledge of each tool capabilities and limitations. This allows to use CAIN for service prototyping (as in currently being done in the EU IST aceMedia Integrated Project), service deployment (reducing the flexibility and numbers of CATs), and as a benchmarking framework. For the later, it will be necessary to incorporate the modules required for providing as result of the adaptation a report of resource consumption (CAT processing time, CPU and memory use, power consumed, temporary storage space, …) as well as modules for comparing the quality of different versions of adapted content, both being the same or different adaptations, taking into account not only objective, but also subjective quality measures focus in the user’s perceived quality even in the case of transmoding.
This work is partially supported by the European Commission 6th Framework Program under project FP6-001765 (aceMedia). This work is also supported by the Ministerio de Ciencia y Tecnología of the Spanish Government under project TIN2004-07860 (MEDUSA) and by the Comunidad de Madrid under project P-TIC-0223-0505 (PROMULTIDIS). The authors want to thank Víctor Fernández-Carbajales for successful testing of the extensibility mechanism.