 |
 |
 |
 |
|
Seamless Media Adaptation
|
 |
|
 |
| In future multimedia communication
scenarios, different aspects of heterogeneity will pose several technical
challenges for transmission technologies. Typical applications will serve
large groups of users with particular Quality-of-Service (QoS) wishes, utilizing
a broad range of heterogeneous multimedia devices connected via different
(and probably in most cases wireless) access network technologies. In order
to save bandwidth, maximize subjective perceived quality and allow for simple
and secure implementations of network nodes, efficient adaptive media transport
mechanisms will become necessary. Adaptation can be performed at various places. At the application layer, specific requirements are well-known, but the mechanisms have to be 're-invented' for every application again and also interoperability is not ensured. Adaptive operating systems have a global view on the nodes resource situation and could allows for optimized utilization and fairness, but the application semantic is not known. Also, operability between different operating systems is hard to achieve. We propose MediaStorm as an adaptive middleware layer between applications and operating systems, which can combine both advantages. This allows for an efficient and fair handling of all involved media streams according to the overall resource situations. |
 |
|
MediaStorm: Adaptive streaming middleware
|
| MediaStorm is a Java-based multimedia streaming library, which abstracts from the underlying multmedia system (e.g. Apple Quicktime, Sun Java Media Framework for J2SE, Sun Java MMAPI for J2ME, etc.) and operating system. Easy interfaces allow for a simplified development of feature-rich adaptive multimedia applications (like video-on-demand, IP-telephony, Internet Radio, and many more). |
 |
|
End-to-end Coverage
|
| MediaStorm can be installed on any node of the transmission scheme and is able to automatically perform automatic seamless media adaptation mechanisms. |
 |
 |
In a master thesis project
in cooperation with the University of Stuttgart we analyzed the perceived
quality of different audio codecs (e.g. DVI, uLaw, G.723, GSM, MP-3, etc.).
The human ear is especially sensitive to loss in audio data. Therefore,
we developed an adaptation strategy, which is able to reduce loss significantly
by switching to the most appropriate codec regarding the current transmission
characteristics. |
|
Adaptive Audio Codec Switching
|
|
 |
In a master thesis projects
together with the University of Applied Sciences in Mannheim, we developed
a new Java native interface for the Sun Java Media Framework (JMF) to support
realtime video filtering and compression attribute adaptation using a DivX-based
MPEG-4 codec. By using a complex adaptation algorithm, we were able to control
the intra-frame quality and the actual frame rate of an MPEG-4 stream during
the transmission to adapt to the current network situation (indicated by
loss and jitter values). By using a network simulator, the performance of
the algorithm for different load situations was tested succesfully. In order
to add further support for network handoff situations we included a pre-handoff
adaptation strategy, improving the results even further. In a second master thesis project with the University of Applied Sciences in Mannheim, we developed JMF RTP packetizer, depacketizer and filter modules for the WaveVideo codec developed at the ETH Zurich. This allowed us to test the impact of different filters (e.g. Quality, Datarate, Framerate) on the quality of wavelet coded video material. The system allows for multicast streaming and enables automatic sender datarate adaptation based on receiver feedbacks. See publication reference below for more information about our video filtering approaches. |
WaveVideo Example with varying Quality (1.0, 0.8,
0.75, 0.5, 0.25, 0.15) |
|
 |
A media processing chain
is often a complicated collection of codecs, packet handlers, media filters,
and memory buffers for processing media data. Adapting such a media processing
chain often requires a lengthy process of deconstructing parts of chain
or the complete chain, and rebuilding a replacement. This process is impacted by two types of delay, present within every element of the media processing chain - setup delay and intrinsic delay. Setup delay is defined as the time required by the processing element to initialise its internal data structures, acquire necessary resources and become ready for operation. Intrinsic delay is characterised as the additional time, beyond setup delay, required by a processing element to produce a particular output with a given input. Together, setup and intrinsic delays can introduce additional levels of data loss and extended delays that can degrade quality. Conventional adaptation mechanisms utilize a simplified method of deconstructing and reconstructing media processing chains without regard for either setup or intrinsic delays. Since these delays can be significant and cause data loss, we developed a new seamless method that avoids these effects. |
 |
|
|
Parallel Media Processing Chains
|
|
| By using simultaneous processing chains instead of sequential
teardown and setup phases, the adaptation process has been significantly
improved. We proved our system with an analytical model and also build a
Java-based prototype implementation. |
|
 |
|
|
Example application for seamless
Adapter (Internet Television)
|
|
| Switching between video codecs for example can be realized
with a gap time below 1ms and no loss at all. See publication reference below for more information. |
 |
|
|
 |
|
For more information please use our contact form. Please see also the related MASA
QoS project page. |
| Last modified 02-Jun-2005 |
| Back
to Projects | Top |
|
|
 |
|
| © Network Laboratories Heidelberg 2004-2008 | ||
Christian Kücherer,
Quality of Service with Adaptive Video Coding,
Diplomarbeit im Studiengang Informatik, 2001, University of Applied Sciences
Mannheim.