Principles of DVC - Distributed Multiple Description Coding - page 44

1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

n=25534
n=6336
n=396
Slepian-Wolf bound

0.1

0

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

H(X|Y)

Fig. 3.7 Rate curves of Slepian-Wolf codec based on turbo coding

5. Do joint decoding; we get the corresponding X 0 from Z after turbo encoder. If
X 0 meets the requirements of rate distortion or it reaches the maximum number
of iterations, finish the encoding and give the present used bit rate, otherwise go
back to step 3.

Figure 3.7 shows that when X and Y are under different correlations, the perfor-
mance curve of the Slepian-Wolf encoder is based on RCPT code, and compare it
with the ideal limit value H.XjY/D H.P/ DP log 2 P .1 P/ log 2 .1 P/ , P
is the cross-transition probability. As can be seen from the figure, RCPT can realize
bit rate adaptive; it has a better compression effect when the transition probability P
is low; the gap with the ideal value is about 0.1 dB. But when the correlation is poor
(when P is large) and the gap with ideal boundary value reaches more than 0.2 dB,
for example, when P>0:2 , we can only use R X D H.X/ D 1 restore X ,almost
no compression performance; the reason is that drill based on the parity bit has a
great impact on the performance of turbo. The Slepian-Wolf encoder with higher
performance requires improvement from the structure of turbo codes; in addition,
the experiment shows that with the increasing of the input sequence length, due to
the fact that the capacity to correct is stronger when the turbo sequence is long, the
performance of Slepian-Wolf is closer to the Slepian-Wolf limit.