In this dissertation, we studied two seemingly uncorrelated, but conceptually
related problems in terms of source-channel coding: 1) wireless image transmission
and 2) Costa ("dirty-paper") code design.
In the first part of the dissertation, we consider progressive image transmission
over a wireless system employing space-time coded OFDM. The space-time coded
OFDM system based on a newly built broadband MIMO fading model is theoretically
evaluated by assuming perfect channel state information (CSI) at the receiver for
coherent detection. Then an adaptive modulation scheme is proposed to pick the
constellation size that offers the best reconstructed image quality for each average
signal-to-noise ratio (SNR).
A more practical scenario is also considered without the assumption of perfect
CSI. We employ low-complexity decision-feedback decoding for differentially space-
time coded OFDM systems to exploit transmitter diversity. For JSCC, we adopt a
product channel code structure that is proven to provide powerful error protection and
bursty error correction. To further improve the system performance, we also apply
the powerful iterative (turbo) coding techniques and propose the iterative decoding
of differentially space-time coded multiple descriptions of images.
The second part of the dissertation deals with practical dirty-paper code designs. We first invoke an information-theoretical interpretation of algebraic binning and
motivate the code design guidelines in terms of source-channel coding. Then two
dirty-paper code designs are proposed. The first is a nested turbo construction based
on soft-output trellis-coded quantization (SOTCQ) for source coding and turbo trellis-
coded modulation (TTCM) for channel coding. A novel procedure is devised to
balance the dimensionalities of the equivalent lattice codes corresponding to SOTCQ
and TTCM. The second dirty-paper code design employs TCQ and IRA codes for
near-capacity performance. This is done by synergistically combining TCQ with IRA
codes so that they work together as well as they do individually. Our TCQ/IRA
design approaches the dirty-paper capacity limit at the low rate regime (e.g., < 1:0
bit/sample), while our nested SOTCQ/TTCM scheme provides the best performs so
far at medium-to-high rates (e.g., >= 1:0 bit/sample). Thus the two proposed practical
code designs are complementary to each other.
| Identifer | oai:union.ndltd.org:tamu.edu/oai:repository.tamu.edu:1969.1/4800 |
| Date | 25 April 2007 |
| Creators | Sun, Yong |
| Contributors | Xiong, Zixiang |
| Publisher | Texas A&M University |
| Source Sets | Texas A and M University |
| Language | en_US |
| Detected Language | English |
| Type | Book, Thesis, Electronic Dissertation, text |
| Format | 1602539 bytes, electronic, application/pdf, born digital |
Page generated in 0.002 seconds