| |
FlexCode Public Seminar, 16.06.2008, 13.30-15.00,
Nokia's building Hermia 5, Auditorium A156, Visiokatu 1, Tampere, Finland |
|
| |
Title: Turbo
Source Coding
Speaker: Laurent Schmalen
Abstract:
During the last decade, it has been found that channel codes
employing iterative decoders like, e.g., Turbo or LDPC codes,
allow near-Shannon limit transmission of binary data with
moderate computational complexity. It can be shown that these
codes can also be effectively used for lossless or near-lossless
source coding. In our presentation we will first review the
basics of iteratively decodable codes and show then how these
codes can be used for source coding. Afterwards, the concept of
soft decision source decoding is presented which enables the
exploitation of the statistical properties of the source in
order to generate extrinsic information necessary for the use in
iterative receivers. It is finally shown how the soft decision
source decoder can be utilized in an iterative framework in
order to realize near-lossless source coding. It is furthermore
shown that the entropy of the source can be closely reached and
that the system automatically incorporated channel coding and
becomes very flexible in varying channel conditions.
Download PDF version of the slides
|
|
| |
Title: Flexible Multiple Description Coding of Audio
Speaker: Janusz Klejsa
Abstract:
Multiple-description coding (MDC) addresses the ubiquitous problem of packet loss in packet networks by exploiting diversity. The capacity and packet loss rate of modern communication networks is highly variable, both in time and between networks. Existing MDC methods generally do not reflect this variability: the design of an MDC for a particular rate and packet loss environment requires off-line optimization and is, therefore, not suitable for a flexible coder. Moreover the application of MDC to an audio coding scheme that employs signal modeling introduces yet another problem of finding the optimal rate allocation between the model and the descriptions.
This talk will focus on the application of a flexible MDC in a scalable audio coder. Our objective is a coding scheme that adapts in an on-line fashion to a specified rate and packet-loss probability. This includes finding the optimal rate allocation between the model and the descriptions. The goal for the MDC is to react to changing channel conditions by adapting the level of redundancy. The rate allocation problems are treated analytically, leading to a flexible multiple-description audio coder.
Download PDF version of the slides
|
|
|
