EECS CSPL SEMINAR SERIES
FALL TERM 1999


Tom Fuja

Prof. Tom Fuja

Electrical Engineering Department

University of Notre Dame



Thursday, December 9
4:30 - 5:30 P.M.
Room 1311 EECS


Joint source-channel coding: The view from the decoder

Abstract-
"Joint source-channel coding" is a catchall phrase to describe techniques in which the compression function and the error control function in a communication system are combined in some way. One problem with joint source-channel coding - apart from the fact that Claude Shannon said we shouldn't do it - is that it may sacrifice the modularity that is so convenient in modern communications. When we compress a signal, we don't want to have to worry about what kind of channel it will be communicated over or stored in; indeed, when we compress it, we may not know how or where it will be transmitted.

However, if we carry out the source encoding and the channel encoding independently but permit the decoders to cooperate, the loss of modularity is less intrusive. The research described in this talk addresses methods for carrying out joint source-channel decoding based on maximum a posteriori (MAP) principles. The focus is on two kinds of applications: robust transmission of very low bit rate speech, and joint decoding of sources that have been compressed with variable length source codes.



Biosketch-
Tom Fuja received his undergraduate education at the University of Michigan and his graduate education at Cornell University. He spent eleven years on the faculty of the University of Maryland, and in 1997-1998 he was program director for communications research at the National Science Foundation. In 1998 he joined the faculty of the University of Notre Dame, where he is an associate professor of electrical engineering.


return to CSPL Seminars homepage