Performance of MDC and SDC in slowly varying bursty channels

Test Conditions

Source: Larry (QCIF) sequence with 1689 frames at 15 fps

Source Coding: MDC and SDC for comparison.

Channel Coding/Modulation: BCH code/BPSK

Two-state Markov channels: BER in good state is fixed at 0.001. Channel transition possible at the start of each video frame.
Rayleigh channels: two channels at carrier frequencies of 900MHz and 1800MHz, respectively.

Definition of SDC1 and SDC2

We use 2 SDC schemes for comparison with MDC.
SDC1: We code at the same source rate as two descriptions in MDC and use the same FEC for atoms so that total rate is the same.
SDC2: We code at such source rate that PSNR is same as two description MDC with no channel error, and choose FEC to match the overall rate with MDC.

Comparisons in the literature are typically between MDC and SDC1 over lossy channels, showing that MDC outperforms SDC1 at large loss rates and vice versa at small loss rates due to redundancy in MDC. We believe a more interesting comparison would be between MDC and SDC2, both having same error free performance. SDC2 allocates all available redundancy to FEC while MDC uses weaker FEC and allocates remaining redundancy to source coding. Comparison between MDC and SDC2 would test the efficiency of a MDC design.


Two-state Markov channels


Performance of SDC and MDC schemes in two-state Markov channels with different state transition probabilities. (a) slowly varying channels (p1 = 0.05, p2 = 0.05) (b) fast varying channels (p1 = 0.8, p2 = 0.8)
p1 is the channel transition probability from good state to bad state, and p2 is the channel transition probability from bad state to good state,

Rayleigh fading channels




Performance of SDC and MDC schemes in slowly varying Rayleigh fading channels with different interleaving delays. Values of f_dT (product of Doppler frequency and bit interval) for two Rayleigh channels are 5x10^{-6} and 10^{-5}. Interleaving delays are (a) 0 ms (b) 50 ms (c) 100 ms (d) 150 ms