The delay-distortion performance of the position-value based resource allocation is compared in this section with traditional layer-based resource allocation in mobile WiMAX. The parameters of the simulation study are listed as follows. Default time frame duration T = 0.004 second, default channel BER is 0.0001, MAC header is 6 bytes, and fragmentation subheader is 2 bytes. Frequency bandwidth is 20 MHz, and 64-QAM with 3/4 coding rate and 1/4 cyclic prefix are used.
Figures 1 and 2 indicate the average loss ratio and expected delay trade-off for delivering a typical SDU with 1400 bytes using different fragmentation thresholds and SR-ARQ retry limit strategies. From these figures it is clear to see, with larger fragmentation number (lower fragmentation threshold and shorter PDU length accordingly) and higher SR-ARQ retry limit, the SDU packet loss ratio is decreased considerably. However, the penalty of such packet loss ratio decreasing is the prolonged delay of successful SDU delivery, mainly because of the retransmission latency. This is because mobile WiMAX is basically TDMA based scheduling, retransmission has to be reissued in the next frame duration. Thus, quality and latency form the trade-off that can be fine-tuned in mobile WiMAX transmission strategies optimization.
Figure 1: SDU loss ratio for an application layer packet with 1400-byte length, at channel BER 1e-4.
Figure 2: SDU delay expectation for an application layer packet with 1400-byte length, at channel BER 1e-4.
Figure 3 depict the delay-distortion performance comparison of the position-value based approach and layer-based approaches. For both of these approaches, image qualities with loose delay constraints are better than those with strict delay constraints. This is because with loose delay constraints, more network resource especially the SR-ARQ retransmissions can be allocated to the code stream, which improves the packet delivery ratio and thus the picture quality considerably. The position-value based approach achieves better delay-distortion performance than layer-based approach with the same latency budget constraint. Layer-based UEP approaches allocate resource according to the importance of different layers in code stream, and important layers containing coarse image information are more effectively protected while unimportant layers containing imagefine details are less protected. The position oriented approach allocates resource more efficiently by considering not only different layers’ unequal importance, but also the unequal importance of position and value information in each layer. With the position-value based resource allocation, the p-segments especially those in the coarse image quality layers are more effectively protected to improve image quality; and the v-segments especially those in fine details enhancement layers are less protected to reduce delay penalty. From this figure we can see, with 1e-4 channel BER the position-value based approach shows quality improvement up to 7–8 dB in terms of PSNR over traditional layer based approach with the same delay constraint. In the worst cases, i.e., with ultra tight or ultra loose delay constraints, the position-based approach has similar performance as layer based approach. This is because it has either not enough network resource or over excessive network resource for allocation. In those situations, the performance is not confined by efficiency of resource allocation, but the amount of network resource itself.
Figure 3: Image quality and delay-bound for different resource allocation schemes at BER 1e-4.
