[CPCC] RESCHEDULED PRESENTATION: Joint Source/Channel Decoding with Factor Graphs

[Ender Ayanoglu]

	Joint Source Channel Decoding with Factor Graphs


			*** NOTE NEW DATE ***


			     Inanc Inan

			    UCI/EECS/CPCC

			 November 5, Wednesday
 			 *** NOTE NEW DATE ***

				11 AM
			 Engineering Tower 331


By exploiting the residual redundancy of the source coded data
throughout channel decoding, not only the performance of conventional
system with separate complex encoders can be achieved with lower
complexity encoding, but also significant decoding gains with respect
to conventional separate decoding can be obtained. However, joint source
channel decoding (JSCD) approach is more complex than the separation-based
approach, thus it calls for a new and efficient framework that exposes
joint design gains at a reasonable complexity.

The proposed factor graph framework provides a computationally efficient
way of combining source and channel decoding using iterative methods such
as the sum-product algorithm. Alternative factor graphs for JSCD are
obtained for systems with fixed and variable length source codes and
low density parity check (LDPC) channel codes. Although the proposed
factor graphs have short and tight loops, JSCD with the sum-product
algorithm shows promising performance in simulations for all the factor
graphs except one.

Experiment results also show that JSCD performance is improved by
increasing residual redundancy of the source coded data, decreasing
channel code rate, and increasing channel code blocklength. These
results lead to a JSCD performance trade-off between compression rate
at the source encoder and channel code rate at the channel encoder. It
is shown that JSCD may perform better decoding for redundant data coded
with a higher rate channel code than less redundant data coded with a
lower rate channel code.


Speaker's Biography: Inanc Inan received his M.S. degree in Electrical
Engineering from Bilkent University on September 2003. He is currently
a graduate student researcher at UCI. The work he will present is his
recent M.S. thesis.

Directions: http://www.eng.uci.edu/cpcc/?page=directions