A model is proposed for replacing the video codec in networking studies, which is tested for sources with a mean bitrate in the range of 10Kb/s to 10Mb/s. The standard frame sizes investigated are the International Telecommunications Union's (ITU) 352 x 288-pixel Common Intermediate Format (CIF), 176 x 144-pixel Quarter CIF (QCIF), 704 x 576-pixel 4CIF and 128 x 96-pixel Sub-QCIF (SQCIF). The proposed 20-state 'oscillation-scaled' Markov model was found to represent the relevant video codec characteristics adequately. The performance of this model is then tested in the context of a novel Multi-frame Packet Reservation Multiple Access or MF-PRMA scheme, which i s s h o wn in Figure 15 to outperform conventional PRMA in terms of both its delay and packet dropping performance, reaching throughput values as high as 79 %.
MOTIVATION AND BACKGROUNDIn recent years there have been increased worldwide research activities in the area of modelling video codecs, mainly for studying the performance of multiple access (MAC) schemes 1]-8], where a simple, but su ciently accurate video source model is needed. This is motivated by the fact recognised by researchers that in studying the performance of various MAC schemes it is unnecessary and extremely time-consuming to conduct bittrue simulations of the physical layer, including video compression, channel coding and modulation, as long as these physical layer operations can be adequately modelled by an appropriate source model. This is particularly so for example for high-rate, The motivation of this contribution is two-fold. Firstly, a n o vel video model is proposed and its ability to characterise various video sources is studied in various practical scenarios. The proposed source-modelling technique was investigated in the context of the H.263 video codec 10, 9], since this is to date the most successful standard video codec. Although this codec extensively invokes errorsensitive run-length coding techniques for achieving maximum coding e ciency, since it was designed for low errorrate Gaussian channels, in recent years a variety of techniques have been proposed for facilitating its employment in wireless communications. In particular, Khansari, Jalali, Dubois and Mermelstein from INRS in Canada investigated the employment of the H.261 codec, which was the predecessor of the H.263 scheme 11], while F arber, Steinbach and Girod 12] at Erlangen University in Germany a s well as Cherriman and Hanzo in the UK 13]-17] considered various H.263-based video systems. It was found that if the video packets can be superimposed a simple acknowledgement ag in both directions for informing the local and remote encoder of the success or failure of the transmitted packet, the contents of the local and remote reconstruction frame bu ers can remain identical, since corrupted packets can be dropped at both ends. This resulted in an extremely robust operation, tolerating packet dropping rates as high as 5%
1The video frame areas represented by the dropped packets can then be updated in ...
