Joint Source-Channel Coding Fundamentals by Tor A. Ramstad

Joint source-channel coding (JSCC) is promising for practical delays and complexities, as opposed to Shannon's separation theorem which states that source and channel coding can be done separately, but in general
at the cost of infinite delay and complexity.Our approach to JSCC is based on direct mappings (Shannon-Kotelnikov mappings) from the source, represented by its samples, to channel samples, thus avoiding any form of bit representation.We will first review "optimal performance theoretically attainable" (OPTA), which imposes bandwidth stretching between source and channel signals. But this also leads us to a way to optimize linear systems. The
most general case considered consists of subband decomposition, reshuffling of channels and finally OFDM transmission.We then introduce the principles of nonlinear mappings. Some optimized mappings will be shown, and we will demonstrate a system for image transmission based on subband decomposition and use of several dimension
changing mappings which are allocated according to the necessary bandwidth change among the components.Bio: Professor Tor A. Ramstad received his MSc and PhD degrees from the Norwegian Institute of Technology (now the Norwegian University of Science and Technology (NTNU)) in 1968 and 1971, respectively. He
became a full professor in 1983 at what is now called the Faculty of Informatics, Mathematics and Electrical Engineering. He is presently a visiting scholar at University of California, Santa Barbara.His research interests include multirate signal processing, speech and image processing with emphasis on image and video communications, where joint source-channel coding is central. Low power communication for medical
sensor networks is presently the main research topic.Host: Dr. Sanjit Mitra

Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248

Audiences: Everyone Is Invited

Contact: Gloria Halfacre